diff --git a/src/valuesets/enums/__init__.py b/src/valuesets/enums/__init__.py index d0c94158..2f53d17f 100644 --- a/src/valuesets/enums/__init__.py +++ b/src/valuesets/enums/__init__.py @@ -89,6 +89,7 @@ from .computing.croissant_ml import MLDataType, DatasetEncodingFormat, DatasetSplitType, MLLicenseType, MLFieldRole, CompressionFormat, MLMediaType, MLModalityType from .computing.data_centers import DataCenterCoolingType, DataCenterTierLevel from .computing.file_formats import ImageFileFormatEnum, DocumentFormatEnum, DataFormatEnum, ArchiveFormatEnum, VideoFormatEnum, AudioFormatEnum, ProgrammingLanguageFileEnum, NetworkProtocolEnum +from .computing.frontier_ai_models import FrontierModelEnum from .computing.geospatial_formats import GeospatialRasterFormat, GeospatialVectorFormat from .computing.maturity_levels import TechnologyReadinessLevel, SoftwareMaturityLevel, CapabilityMaturityLevel, StandardsMaturityLevel, ProjectMaturityLevel, DataMaturityLevel, OpenSourceMaturityLevel from .computing.microelectronics import SemiconductorMaterialType, ChipFabricationNodeType @@ -461,6 +462,7 @@ "FossilFuelTypeEnum", "FraudDetectionEnum", "FrequencyUnitEnum", + "FrontierModelEnum", "FuelAssemblyTypeEnum", "FuelCycleFacilityTypeEnum", "FuelCycleOperationEnum", diff --git a/src/valuesets/enums/computing/frontier_ai_models.py b/src/valuesets/enums/computing/frontier_ai_models.py new file mode 100644 index 00000000..27aecadd --- /dev/null +++ b/src/valuesets/enums/computing/frontier_ai_models.py @@ -0,0 +1,132 @@ +""" +Frontier AI Model Value Sets + +A hierarchical value set of frontier large language / foundation models, organized as provider > model family > specific model release. Leaf nodes use the canonical model identifier as used in the providers' APIs and "/model" selectors (e.g. claude-opus-4-8, gpt-5.5, gemini-3.5-flash). Intermediate nodes group models by developer and family. Per-model metadata (knowledge/training cutoff, context window, max output tokens, release date, links to model cards) is attached as annotations and see_also links where available. +NOTE: The frontier model landscape changes rapidly; values reflect the state of provider documentation as of mid-2026. Canonical API identifiers for open-weight models (Llama, DeepSeek, Mistral, Kimi) can vary between inference hosts; the identifiers given are the developer's reference names. + +Generated from: computing/frontier_ai_models.yaml +""" + +from __future__ import annotations + +from valuesets.generators.rich_enum import RichEnum + +class FrontierModelEnum(RichEnum): + """ + Hierarchical enumeration of frontier AI models. The hierarchy has three tiers: developer (e.g. ANTHROPIC), model family (e.g. CLAUDE_OPUS), and specific model release identified by its canonical API model id (e.g. claude-opus-4-8). Intermediate developer and family nodes are uppercase categorical values; leaf nodes are the verbatim API identifiers. + """ + # Enum members + ANTHROPIC = "ANTHROPIC" + CLAUDE_OPUS = "CLAUDE_OPUS" + CLAUDE_OPUS_4_8 = "claude-opus-4-8" + CLAUDE_OPUS_4_7 = "claude-opus-4-7" + CLAUDE_OPUS_4_6 = "claude-opus-4-6" + CLAUDE_OPUS_4_5 = "claude-opus-4-5" + CLAUDE_SONNET = "CLAUDE_SONNET" + CLAUDE_SONNET_4_6 = "claude-sonnet-4-6" + CLAUDE_SONNET_4_5 = "claude-sonnet-4-5" + CLAUDE_HAIKU = "CLAUDE_HAIKU" + CLAUDE_HAIKU_4_5 = "claude-haiku-4-5" + CLAUDE_FABLE = "CLAUDE_FABLE" + CLAUDE_FABLE_5 = "claude-fable-5" + CLAUDE_MYTHOS = "CLAUDE_MYTHOS" + CLAUDE_MYTHOS_5 = "claude-mythos-5" + OPENAI = "OPENAI" + GPT_5 = "GPT_5" + GPT_5_5 = "gpt-5.5" + GPT_5_4 = "gpt-5.4" + GPT_5_4_MINI = "gpt-5.4-mini" + GPT_5_2 = "gpt-5.2" + GPT_5_2_CODEX = "gpt-5.2-codex" + GOOGLE = "GOOGLE" + GEMINI_3 = "GEMINI_3" + GEMINI_3_5_FLASH = "gemini-3.5-flash" + GEMINI_3_1_PRO_PREVIEW = "gemini-3.1-pro-preview" + GEMINI_3_FLASH_PREVIEW = "gemini-3-flash-preview" + GEMINI_3_1_FLASH_LITE = "gemini-3.1-flash-lite" + GEMINI_2_5 = "GEMINI_2_5" + GEMINI_2_5_PRO = "gemini-2.5-pro" + GEMINI_2_5_FLASH = "gemini-2.5-flash" + GEMINI_2_5_FLASH_LITE = "gemini-2.5-flash-lite" + XAI = "XAI" + GROK_4 = "GROK_4" + GROK_4_3 = "grok-4.3" + GROK_4_20_0309_REASONING = "grok-4.20-0309-reasoning" + GROK_4_0709 = "grok-4-0709" + GROK_CODE = "GROK_CODE" + GROK_CODE_FAST_1 = "grok-code-fast-1" + META = "META" + LLAMA_4 = "LLAMA_4" + LLAMA_4_MAVERICK = "llama-4-maverick" + LLAMA_4_SCOUT = "llama-4-scout" + DEEPSEEK = "DEEPSEEK" + DEEPSEEK_MODELS = "DEEPSEEK_MODELS" + DEEPSEEK_CHAT = "deepseek-chat" + DEEPSEEK_REASONER = "deepseek-reasoner" + MISTRAL = "MISTRAL" + MISTRAL_LARGE = "MISTRAL_LARGE" + MISTRAL_LARGE_3 = "mistral-large-3" + MOONSHOT = "MOONSHOT" + KIMI = "KIMI" + KIMI_K2_6 = "kimi-k2.6" + +# Set metadata after class creation +FrontierModelEnum._metadata = { + "ANTHROPIC": {'description': 'Models developed by Anthropic (the Claude family)', 'annotations': {'node_type': 'developer', 'homepage': 'https://www.anthropic.com'}}, + "CLAUDE_OPUS": {'description': "Claude Opus - Anthropic's most capable Opus-tier model family", 'annotations': {'node_type': 'family', 'tier': 'most capable'}}, + "CLAUDE_OPUS_4_8": {'description': "Claude Opus 4.8 - Anthropic's most capable Opus-tier model; highly autonomous, state-of-the-art on long-horizon agentic work, knowledge work, and memory. Adaptive thinking only.", 'annotations': {'node_type': 'model', 'api_model_id': 'claude-opus-4-8', 'developer': 'Anthropic', 'context_window': 1000000, 'max_output_tokens': 128000, 'input_usd_per_mtok': 5.0, 'output_usd_per_mtok': 25.0}, 'aliases': ['Opus 4.8', 'opus']}, + "CLAUDE_OPUS_4_7": {'description': 'Claude Opus 4.7 - previous-generation Opus; highly autonomous, strong on long-horizon agentic work, vision, and memory. Adaptive thinking only.', 'annotations': {'node_type': 'model', 'api_model_id': 'claude-opus-4-7', 'developer': 'Anthropic', 'context_window': 1000000, 'max_output_tokens': 128000}, 'aliases': ['Opus 4.7']}, + "CLAUDE_OPUS_4_6": {'description': 'Claude Opus 4.6 - older Opus model supporting adaptive thinking and 128K max output tokens.', 'annotations': {'node_type': 'model', 'api_model_id': 'claude-opus-4-6', 'developer': 'Anthropic', 'context_window': 1000000, 'max_output_tokens': 128000}, 'aliases': ['Opus 4.6']}, + "CLAUDE_OPUS_4_5": {'description': 'Claude Opus 4.5 - legacy Opus model (still active).', 'annotations': {'node_type': 'model', 'api_model_id': 'claude-opus-4-5', 'full_model_id': 'claude-opus-4-5-20251101', 'developer': 'Anthropic'}, 'aliases': ['Opus 4.5']}, + "CLAUDE_SONNET": {'description': "Claude Sonnet - Anthropic's balanced speed / intelligence family", 'annotations': {'node_type': 'family', 'tier': 'balanced'}}, + "CLAUDE_SONNET_4_6": {'description': "Claude Sonnet 4.6 - Anthropic's best combination of speed and intelligence. Supports adaptive thinking.", 'annotations': {'node_type': 'model', 'api_model_id': 'claude-sonnet-4-6', 'developer': 'Anthropic', 'context_window': 1000000, 'max_output_tokens': 64000, 'input_usd_per_mtok': 3.0, 'output_usd_per_mtok': 15.0}, 'aliases': ['Sonnet 4.6', 'sonnet']}, + "CLAUDE_SONNET_4_5": {'description': 'Claude Sonnet 4.5 - legacy Sonnet model (still active).', 'annotations': {'node_type': 'model', 'api_model_id': 'claude-sonnet-4-5', 'full_model_id': 'claude-sonnet-4-5-20250929', 'developer': 'Anthropic'}, 'aliases': ['Sonnet 4.5']}, + "CLAUDE_HAIKU": {'description': "Claude Haiku - Anthropic's fastest and most cost-effective family", 'annotations': {'node_type': 'family', 'tier': 'fast'}}, + "CLAUDE_HAIKU_4_5": {'description': "Claude Haiku 4.5 - Anthropic's fastest and most cost-effective model for simple, latency-sensitive tasks.", 'annotations': {'node_type': 'model', 'api_model_id': 'claude-haiku-4-5', 'full_model_id': 'claude-haiku-4-5-20251001', 'developer': 'Anthropic', 'context_window': 200000, 'max_output_tokens': 64000, 'input_usd_per_mtok': 1.0, 'output_usd_per_mtok': 5.0}, 'aliases': ['Haiku 4.5', 'haiku']}, + "CLAUDE_FABLE": {'description': "Claude Fable - Anthropic's most capable widely released model family, for the most demanding reasoning and long-horizon agentic work.", 'annotations': {'node_type': 'family', 'tier': 'frontier'}}, + "CLAUDE_FABLE_5": {'description': "Claude Fable 5 - Anthropic's most capable widely released model. Thinking is always on; raw chain of thought is never returned. Requires 30-day data retention.", 'annotations': {'node_type': 'model', 'api_model_id': 'claude-fable-5', 'developer': 'Anthropic', 'context_window': 1000000, 'max_output_tokens': 128000, 'input_usd_per_mtok': 10.0, 'output_usd_per_mtok': 50.0}, 'aliases': ['Fable 5', 'fable']}, + "CLAUDE_MYTHOS": {'description': 'Claude Mythos - same capabilities and API behavior as Claude Fable, available exclusively through Project Glasswing.', 'annotations': {'node_type': 'family', 'tier': 'frontier', 'availability': 'Project Glasswing only'}}, + "CLAUDE_MYTHOS_5": {'description': 'Claude Mythos 5 - same capabilities, pricing, limits, and API behavior as Claude Fable 5; available only through Project Glasswing.', 'annotations': {'node_type': 'model', 'api_model_id': 'claude-mythos-5', 'developer': 'Anthropic', 'context_window': 1000000, 'max_output_tokens': 128000}, 'aliases': ['Mythos 5']}, + "OPENAI": {'description': 'Models developed by OpenAI (the GPT family)', 'annotations': {'node_type': 'developer', 'homepage': 'https://openai.com'}}, + "GPT_5": {'description': "GPT-5 - OpenAI's frontier model family for coding and professional work", 'annotations': {'node_type': 'family', 'tier': 'frontier'}}, + "GPT_5_5": {'description': 'GPT-5.5 - OpenAI\'s newest frontier model, "a new class of intelligence for coding and professional work".', 'annotations': {'node_type': 'model', 'api_model_id': 'gpt-5.5', 'developer': 'OpenAI', 'training_cutoff': '2025-12-01', 'context_window': 1000000}, 'aliases': ['GPT-5.5']}, + "GPT_5_4": {'description': 'GPT-5.4 - a more affordable model for coding and professional work.', 'annotations': {'node_type': 'model', 'api_model_id': 'gpt-5.4', 'developer': 'OpenAI', 'training_cutoff': '2025-08-31', 'context_window': 1000000}, 'aliases': ['GPT-5.4']}, + "GPT_5_4_MINI": {'description': "GPT-5.4 mini - OpenAI's strongest mini model for coding, computer use, and subagents; lower latency and cost.", 'annotations': {'node_type': 'model', 'api_model_id': 'gpt-5.4-mini', 'developer': 'OpenAI', 'training_cutoff': '2025-08-31', 'context_window': 400000}, 'aliases': ['GPT-5.4 mini']}, + "GPT_5_2": {'description': 'GPT-5.2 - previous frontier model for complex professional work.', 'annotations': {'node_type': 'model', 'api_model_id': 'gpt-5.2', 'developer': 'OpenAI'}, 'aliases': ['GPT-5.2']}, + "GPT_5_2_CODEX": {'description': 'GPT-5.2-Codex - coding-specialized variant for API-authenticated Codex workflows.', 'annotations': {'node_type': 'model', 'api_model_id': 'gpt-5.2-codex', 'developer': 'OpenAI', 'specialization': 'coding'}, 'aliases': ['GPT-5.2 Codex']}, + "GOOGLE": {'description': 'Models developed by Google DeepMind (the Gemini family)', 'annotations': {'node_type': 'developer', 'homepage': 'https://deepmind.google'}}, + "GEMINI_3": {'description': "Gemini 3 - Google's latest generation multimodal model family", 'annotations': {'node_type': 'family', 'tier': 'frontier'}}, + "GEMINI_3_5_FLASH": {'description': "Gemini 3.5 Flash - Google's most intelligent model for agentic and coding tasks, delivering near-Pro intelligence at Flash-tier speed and cost.", 'annotations': {'node_type': 'model', 'api_model_id': 'gemini-3.5-flash', 'developer': 'Google DeepMind', 'release_date': '2026-05-19', 'context_window': 1000000}, 'aliases': ['Gemini 3.5 Flash']}, + "GEMINI_3_1_PRO_PREVIEW": {'description': 'Gemini 3.1 Pro - reasoning-first model optimized for complex agentic workflows and coding, with adaptive thinking and integrated grounding.', 'annotations': {'node_type': 'model', 'api_model_id': 'gemini-3.1-pro-preview', 'developer': 'Google DeepMind', 'context_window': 1000000}, 'aliases': ['Gemini 3.1 Pro']}, + "GEMINI_3_FLASH_PREVIEW": {'description': 'Gemini 3 Flash - frontier-class performance at reduced cost.', 'annotations': {'node_type': 'model', 'api_model_id': 'gemini-3-flash-preview', 'developer': 'Google DeepMind'}, 'aliases': ['Gemini 3 Flash']}, + "GEMINI_3_1_FLASH_LITE": {'description': 'Gemini 3.1 Flash-Lite - most cost-efficient Gemini model, optimized for low latency and high-volume traffic.', 'annotations': {'node_type': 'model', 'api_model_id': 'gemini-3.1-flash-lite', 'developer': 'Google DeepMind'}, 'aliases': ['Gemini 3.1 Flash-Lite']}, + "GEMINI_2_5": {'description': "Gemini 2.5 - Google's previous-generation multimodal model family", 'annotations': {'node_type': 'family'}}, + "GEMINI_2_5_PRO": {'description': 'Gemini 2.5 Pro - high-capability model for complex reasoning and coding with adaptive thinking and a 1M token context.', 'annotations': {'node_type': 'model', 'api_model_id': 'gemini-2.5-pro', 'developer': 'Google DeepMind', 'context_window': 1000000}, 'aliases': ['Gemini 2.5 Pro']}, + "GEMINI_2_5_FLASH": {'description': 'Gemini 2.5 Flash - best price-performance for reasoning tasks.', 'annotations': {'node_type': 'model', 'api_model_id': 'gemini-2.5-flash', 'developer': 'Google DeepMind'}, 'aliases': ['Gemini 2.5 Flash']}, + "GEMINI_2_5_FLASH_LITE": {'description': 'Gemini 2.5 Flash-Lite - fastest and most budget-friendly option.', 'annotations': {'node_type': 'model', 'api_model_id': 'gemini-2.5-flash-lite', 'developer': 'Google DeepMind'}, 'aliases': ['Gemini 2.5 Flash-Lite']}, + "XAI": {'description': 'Models developed by xAI (the Grok family)', 'annotations': {'node_type': 'developer', 'homepage': 'https://x.ai'}}, + "GROK_4": {'description': "Grok 4 - xAI's reasoning-first flagship model family", 'annotations': {'node_type': 'family', 'tier': 'frontier'}}, + "GROK_4_3": {'description': "Grok 4.3 - xAI's reasoning-first flagship; the most intelligent and fastest model xAI has built.", 'annotations': {'node_type': 'model', 'api_model_id': 'grok-4.3', 'developer': 'xAI', 'release_date': '2026-04-30', 'context_window': 1000000}, 'aliases': ['Grok 4.3']}, + "GROK_4_20_0309_REASONING": {'description': 'Grok 4.20 (reasoning) - agentic tool-calling model noted for a very low hallucination rate.', 'annotations': {'node_type': 'model', 'api_model_id': 'grok-4.20-0309-reasoning', 'developer': 'xAI', 'context_window': 1000000}, 'aliases': ['Grok 4.20']}, + "GROK_4_0709": {'description': 'Grok 4 - earlier Grok 4 release.', 'annotations': {'node_type': 'model', 'api_model_id': 'grok-4-0709', 'developer': 'xAI', 'knowledge_cutoff': '2024-11'}, 'aliases': ['Grok 4']}, + "GROK_CODE": {'description': "Grok Code - xAI's dedicated agentic coding model family", 'annotations': {'node_type': 'family', 'specialization': 'coding'}}, + "GROK_CODE_FAST_1": {'description': 'Grok Code Fast 1 - dedicated agentic coding model.', 'annotations': {'node_type': 'model', 'api_model_id': 'grok-code-fast-1', 'developer': 'xAI', 'specialization': 'coding'}, 'aliases': ['Grok Code Fast']}, + "META": {'description': 'Open-weight models developed by Meta (the Llama family)', 'annotations': {'node_type': 'developer', 'homepage': 'https://www.llama.com', 'license': 'Llama Community License'}}, + "LLAMA_4": {'description': "Llama 4 - Meta's open-weight mixture-of-experts model family", 'annotations': {'node_type': 'family'}}, + "LLAMA_4_MAVERICK": {'description': 'Llama 4 Maverick - general-purpose open-weight chat model; highest MMLU among Llama 4 variants.', 'annotations': {'node_type': 'model', 'api_model_id': 'llama-4-maverick', 'developer': 'Meta'}, 'aliases': ['Llama 4 Maverick']}, + "LLAMA_4_SCOUT": {'description': 'Llama 4 Scout - open-weight model supporting very long context (up to 10M tokens).', 'annotations': {'node_type': 'model', 'api_model_id': 'llama-4-scout', 'developer': 'Meta', 'context_window': 10000000}, 'aliases': ['Llama 4 Scout']}, + "DEEPSEEK": {'description': 'Open-weight models developed by DeepSeek', 'annotations': {'node_type': 'developer', 'homepage': 'https://www.deepseek.com'}}, + "DEEPSEEK_MODELS": {'description': 'DeepSeek - open-weight mixture-of-experts chat and reasoning models', 'annotations': {'node_type': 'family'}}, + "DEEPSEEK_CHAT": {'description': 'DeepSeek-V3 (served as deepseek-chat) - open-weight MoE model activating ~37B of 671B parameters per token.', 'annotations': {'node_type': 'model', 'api_model_id': 'deepseek-chat', 'developer': 'DeepSeek'}, 'aliases': ['DeepSeek-V3', 'DeepSeek V3']}, + "DEEPSEEK_REASONER": {'description': 'DeepSeek-R1 (served as deepseek-reasoner) - open-weight reasoning model.', 'annotations': {'node_type': 'model', 'api_model_id': 'deepseek-reasoner', 'developer': 'DeepSeek'}, 'aliases': ['DeepSeek-R1', 'DeepSeek R1']}, + "MISTRAL": {'description': 'Models developed by Mistral AI', 'annotations': {'node_type': 'developer', 'homepage': 'https://mistral.ai'}}, + "MISTRAL_LARGE": {'description': "Mistral Large - Mistral AI's flagship open-weight model family", 'annotations': {'node_type': 'family'}}, + "MISTRAL_LARGE_3": {'description': 'Mistral Large 3 - state-of-the-art open-weight MoE model (41B active / 675B total parameters), 256K context, released under Apache 2.0.', 'annotations': {'node_type': 'model', 'api_model_id': 'mistral-large-3', 'developer': 'Mistral AI', 'release_date': '2025-12', 'context_window': 256000, 'license': 'Apache-2.0'}, 'aliases': ['Mistral Large 3']}, + "MOONSHOT": {'description': 'Models developed by Moonshot AI (the Kimi family)', 'annotations': {'node_type': 'developer', 'homepage': 'https://www.moonshot.ai'}}, + "KIMI": {'description': "Kimi - Moonshot AI's open-weight mixture-of-experts model family", 'annotations': {'node_type': 'family'}}, + "KIMI_K2_6": {'description': 'Kimi K2.6 - frontier open-weight MoE coding model (32B active / 1T total parameters).', 'annotations': {'node_type': 'model', 'api_model_id': 'kimi-k2.6', 'developer': 'Moonshot AI'}, 'aliases': ['Kimi K2.6']}, +} + +__all__ = [ + "FrontierModelEnum", +] \ No newline at end of file diff --git a/src/valuesets/schema/computing/frontier_ai_models.yaml b/src/valuesets/schema/computing/frontier_ai_models.yaml new file mode 100644 index 00000000..1621127e --- /dev/null +++ b/src/valuesets/schema/computing/frontier_ai_models.yaml @@ -0,0 +1,674 @@ +name: valuesets-computing-frontier-ai-models +title: Frontier AI Model Value Sets +description: >- + A hierarchical value set of frontier large language / foundation models, + organized as provider > model family > specific model release. Leaf nodes use + the canonical model identifier as used in the providers' APIs and "/model" + selectors (e.g. claude-opus-4-8, gpt-5.5, gemini-3.5-flash). Intermediate nodes + group models by developer and family. Per-model metadata (knowledge/training + cutoff, context window, max output tokens, release date, links to model cards) + is attached as annotations and see_also links where available. + + NOTE: The frontier model landscape changes rapidly; values reflect the state of + provider documentation as of mid-2026. Canonical API identifiers for + open-weight models (Llama, DeepSeek, Mistral, Kimi) can vary between inference + hosts; the identifiers given are the developer's reference names. +id: https://w3id.org/valuesets/computing/frontier_ai_models +imports: +- linkml:types +prefixes: + linkml: https://w3id.org/linkml/ + valuesets: https://w3id.org/valuesets/ + orcid: https://orcid.org/ + valuesets_meta: https://w3id.org/valuesets/meta/ +default_prefix: valuesets + +slots: + frontier_model: + description: A frontier AI model, its family, or its developer + range: FrontierModelEnum + +enums: + FrontierModelEnum: + title: Frontier AI Model + description: >- + Hierarchical enumeration of frontier AI models. The hierarchy has three + tiers: developer (e.g. ANTHROPIC), model family (e.g. CLAUDE_OPUS), and + specific model release identified by its canonical API model id + (e.g. claude-opus-4-8). Intermediate developer and family nodes are + uppercase categorical values; leaf nodes are the verbatim API identifiers. + status: DRAFT + contributors: + - orcid:0000-0002-6601-2165 + - https://github.com/anthropics/claude-code + instantiates: + - valuesets_meta:ValueSetEnumDefinition + permissible_values: + + ########################################################################## + # Anthropic + ########################################################################## + ANTHROPIC: + description: Models developed by Anthropic (the Claude family) + annotations: + node_type: developer + homepage: https://www.anthropic.com + see_also: + - https://docs.claude.com/en/docs/about-claude/models/overview + + CLAUDE_OPUS: + description: Claude Opus - Anthropic's most capable Opus-tier model family + is_a: ANTHROPIC + annotations: + node_type: family + tier: most capable + + claude-opus-4-8: + description: >- + Claude Opus 4.8 - Anthropic's most capable Opus-tier model; highly + autonomous, state-of-the-art on long-horizon agentic work, knowledge + work, and memory. Adaptive thinking only. + is_a: CLAUDE_OPUS + title: Claude Opus 4.8 + aliases: + - Opus 4.8 + - opus + annotations: + node_type: model + api_model_id: claude-opus-4-8 + developer: Anthropic + context_window: 1000000 + max_output_tokens: 128000 + input_usd_per_mtok: 5.00 + output_usd_per_mtok: 25.00 + see_also: + - https://docs.claude.com/en/docs/about-claude/models/overview + + claude-opus-4-7: + description: >- + Claude Opus 4.7 - previous-generation Opus; highly autonomous, strong on + long-horizon agentic work, vision, and memory. Adaptive thinking only. + is_a: CLAUDE_OPUS + title: Claude Opus 4.7 + aliases: + - Opus 4.7 + annotations: + node_type: model + api_model_id: claude-opus-4-7 + developer: Anthropic + context_window: 1000000 + max_output_tokens: 128000 + + claude-opus-4-6: + description: >- + Claude Opus 4.6 - older Opus model supporting adaptive thinking and 128K + max output tokens. + is_a: CLAUDE_OPUS + title: Claude Opus 4.6 + aliases: + - Opus 4.6 + annotations: + node_type: model + api_model_id: claude-opus-4-6 + developer: Anthropic + context_window: 1000000 + max_output_tokens: 128000 + + claude-opus-4-5: + description: Claude Opus 4.5 - legacy Opus model (still active). + is_a: CLAUDE_OPUS + title: Claude Opus 4.5 + aliases: + - Opus 4.5 + annotations: + node_type: model + api_model_id: claude-opus-4-5 + full_model_id: claude-opus-4-5-20251101 + developer: Anthropic + + CLAUDE_SONNET: + description: Claude Sonnet - Anthropic's balanced speed / intelligence family + is_a: ANTHROPIC + annotations: + node_type: family + tier: balanced + + claude-sonnet-4-6: + description: >- + Claude Sonnet 4.6 - Anthropic's best combination of speed and + intelligence. Supports adaptive thinking. + is_a: CLAUDE_SONNET + title: Claude Sonnet 4.6 + aliases: + - Sonnet 4.6 + - sonnet + annotations: + node_type: model + api_model_id: claude-sonnet-4-6 + developer: Anthropic + context_window: 1000000 + max_output_tokens: 64000 + input_usd_per_mtok: 3.00 + output_usd_per_mtok: 15.00 + + claude-sonnet-4-5: + description: Claude Sonnet 4.5 - legacy Sonnet model (still active). + is_a: CLAUDE_SONNET + title: Claude Sonnet 4.5 + aliases: + - Sonnet 4.5 + annotations: + node_type: model + api_model_id: claude-sonnet-4-5 + full_model_id: claude-sonnet-4-5-20250929 + developer: Anthropic + + CLAUDE_HAIKU: + description: Claude Haiku - Anthropic's fastest and most cost-effective family + is_a: ANTHROPIC + annotations: + node_type: family + tier: fast + + claude-haiku-4-5: + description: >- + Claude Haiku 4.5 - Anthropic's fastest and most cost-effective model for + simple, latency-sensitive tasks. + is_a: CLAUDE_HAIKU + title: Claude Haiku 4.5 + aliases: + - Haiku 4.5 + - haiku + annotations: + node_type: model + api_model_id: claude-haiku-4-5 + full_model_id: claude-haiku-4-5-20251001 + developer: Anthropic + context_window: 200000 + max_output_tokens: 64000 + input_usd_per_mtok: 1.00 + output_usd_per_mtok: 5.00 + + CLAUDE_FABLE: + description: >- + Claude Fable - Anthropic's most capable widely released model family, for + the most demanding reasoning and long-horizon agentic work. + is_a: ANTHROPIC + annotations: + node_type: family + tier: frontier + + claude-fable-5: + description: >- + Claude Fable 5 - Anthropic's most capable widely released model. Thinking + is always on; raw chain of thought is never returned. Requires 30-day + data retention. + is_a: CLAUDE_FABLE + title: Claude Fable 5 + aliases: + - Fable 5 + - fable + annotations: + node_type: model + api_model_id: claude-fable-5 + developer: Anthropic + context_window: 1000000 + max_output_tokens: 128000 + input_usd_per_mtok: 10.00 + output_usd_per_mtok: 50.00 + + CLAUDE_MYTHOS: + description: >- + Claude Mythos - same capabilities and API behavior as Claude Fable, + available exclusively through Project Glasswing. + is_a: ANTHROPIC + annotations: + node_type: family + tier: frontier + availability: Project Glasswing only + + claude-mythos-5: + description: >- + Claude Mythos 5 - same capabilities, pricing, limits, and API behavior as + Claude Fable 5; available only through Project Glasswing. + is_a: CLAUDE_MYTHOS + title: Claude Mythos 5 + aliases: + - Mythos 5 + annotations: + node_type: model + api_model_id: claude-mythos-5 + developer: Anthropic + context_window: 1000000 + max_output_tokens: 128000 + + ########################################################################## + # OpenAI + ########################################################################## + OPENAI: + description: Models developed by OpenAI (the GPT family) + annotations: + node_type: developer + homepage: https://openai.com + see_also: + - https://developers.openai.com/api/docs/models + + GPT_5: + description: GPT-5 - OpenAI's frontier model family for coding and professional work + is_a: OPENAI + annotations: + node_type: family + tier: frontier + + gpt-5.5: + description: >- + GPT-5.5 - OpenAI's newest frontier model, "a new class of intelligence + for coding and professional work". + is_a: GPT_5 + title: GPT-5.5 + aliases: + - GPT-5.5 + annotations: + node_type: model + api_model_id: gpt-5.5 + developer: OpenAI + training_cutoff: "2025-12-01" + context_window: 1000000 + see_also: + - https://developers.openai.com/api/docs/models/gpt-5.5 + + gpt-5.4: + description: GPT-5.4 - a more affordable model for coding and professional work. + is_a: GPT_5 + title: GPT-5.4 + aliases: + - GPT-5.4 + annotations: + node_type: model + api_model_id: gpt-5.4 + developer: OpenAI + training_cutoff: "2025-08-31" + context_window: 1000000 + + gpt-5.4-mini: + description: >- + GPT-5.4 mini - OpenAI's strongest mini model for coding, computer use, + and subagents; lower latency and cost. + is_a: GPT_5 + title: GPT-5.4 mini + aliases: + - GPT-5.4 mini + annotations: + node_type: model + api_model_id: gpt-5.4-mini + developer: OpenAI + training_cutoff: "2025-08-31" + context_window: 400000 + + gpt-5.2: + description: GPT-5.2 - previous frontier model for complex professional work. + is_a: GPT_5 + title: GPT-5.2 + aliases: + - GPT-5.2 + annotations: + node_type: model + api_model_id: gpt-5.2 + developer: OpenAI + + gpt-5.2-codex: + description: >- + GPT-5.2-Codex - coding-specialized variant for API-authenticated Codex + workflows. + is_a: GPT_5 + title: GPT-5.2-Codex + aliases: + - GPT-5.2 Codex + annotations: + node_type: model + api_model_id: gpt-5.2-codex + developer: OpenAI + specialization: coding + + ########################################################################## + # Google DeepMind + ########################################################################## + GOOGLE: + description: Models developed by Google DeepMind (the Gemini family) + annotations: + node_type: developer + homepage: https://deepmind.google + see_also: + - https://ai.google.dev/gemini-api/docs/models + + GEMINI_3: + description: Gemini 3 - Google's latest generation multimodal model family + is_a: GOOGLE + annotations: + node_type: family + tier: frontier + + gemini-3.5-flash: + description: >- + Gemini 3.5 Flash - Google's most intelligent model for agentic and coding + tasks, delivering near-Pro intelligence at Flash-tier speed and cost. + is_a: GEMINI_3 + title: Gemini 3.5 Flash + aliases: + - Gemini 3.5 Flash + annotations: + node_type: model + api_model_id: gemini-3.5-flash + developer: Google DeepMind + release_date: "2026-05-19" + context_window: 1000000 + + gemini-3.1-pro-preview: + description: >- + Gemini 3.1 Pro - reasoning-first model optimized for complex agentic + workflows and coding, with adaptive thinking and integrated grounding. + is_a: GEMINI_3 + title: Gemini 3.1 Pro + aliases: + - Gemini 3.1 Pro + annotations: + node_type: model + api_model_id: gemini-3.1-pro-preview + developer: Google DeepMind + context_window: 1000000 + + gemini-3-flash-preview: + description: Gemini 3 Flash - frontier-class performance at reduced cost. + is_a: GEMINI_3 + title: Gemini 3 Flash + aliases: + - Gemini 3 Flash + annotations: + node_type: model + api_model_id: gemini-3-flash-preview + developer: Google DeepMind + + gemini-3.1-flash-lite: + description: >- + Gemini 3.1 Flash-Lite - most cost-efficient Gemini model, optimized for + low latency and high-volume traffic. + is_a: GEMINI_3 + title: Gemini 3.1 Flash-Lite + aliases: + - Gemini 3.1 Flash-Lite + annotations: + node_type: model + api_model_id: gemini-3.1-flash-lite + developer: Google DeepMind + + GEMINI_2_5: + description: Gemini 2.5 - Google's previous-generation multimodal model family + is_a: GOOGLE + annotations: + node_type: family + + gemini-2.5-pro: + description: >- + Gemini 2.5 Pro - high-capability model for complex reasoning and coding + with adaptive thinking and a 1M token context. + is_a: GEMINI_2_5 + title: Gemini 2.5 Pro + aliases: + - Gemini 2.5 Pro + annotations: + node_type: model + api_model_id: gemini-2.5-pro + developer: Google DeepMind + context_window: 1000000 + + gemini-2.5-flash: + description: Gemini 2.5 Flash - best price-performance for reasoning tasks. + is_a: GEMINI_2_5 + title: Gemini 2.5 Flash + aliases: + - Gemini 2.5 Flash + annotations: + node_type: model + api_model_id: gemini-2.5-flash + developer: Google DeepMind + + gemini-2.5-flash-lite: + description: Gemini 2.5 Flash-Lite - fastest and most budget-friendly option. + is_a: GEMINI_2_5 + title: Gemini 2.5 Flash-Lite + aliases: + - Gemini 2.5 Flash-Lite + annotations: + node_type: model + api_model_id: gemini-2.5-flash-lite + developer: Google DeepMind + + ########################################################################## + # xAI + ########################################################################## + XAI: + description: Models developed by xAI (the Grok family) + annotations: + node_type: developer + homepage: https://x.ai + see_also: + - https://docs.x.ai/developers/models + + GROK_4: + description: Grok 4 - xAI's reasoning-first flagship model family + is_a: XAI + annotations: + node_type: family + tier: frontier + + grok-4.3: + description: >- + Grok 4.3 - xAI's reasoning-first flagship; the most intelligent and + fastest model xAI has built. + is_a: GROK_4 + title: Grok 4.3 + aliases: + - Grok 4.3 + annotations: + node_type: model + api_model_id: grok-4.3 + developer: xAI + release_date: "2026-04-30" + context_window: 1000000 + + grok-4.20-0309-reasoning: + description: >- + Grok 4.20 (reasoning) - agentic tool-calling model noted for a very low + hallucination rate. + is_a: GROK_4 + title: Grok 4.20 (reasoning) + aliases: + - Grok 4.20 + annotations: + node_type: model + api_model_id: grok-4.20-0309-reasoning + developer: xAI + context_window: 1000000 + + grok-4-0709: + description: Grok 4 - earlier Grok 4 release. + is_a: GROK_4 + title: Grok 4 + aliases: + - Grok 4 + annotations: + node_type: model + api_model_id: grok-4-0709 + developer: xAI + knowledge_cutoff: "2024-11" + + GROK_CODE: + description: Grok Code - xAI's dedicated agentic coding model family + is_a: XAI + annotations: + node_type: family + specialization: coding + + grok-code-fast-1: + description: Grok Code Fast 1 - dedicated agentic coding model. + is_a: GROK_CODE + title: Grok Code Fast 1 + aliases: + - Grok Code Fast + annotations: + node_type: model + api_model_id: grok-code-fast-1 + developer: xAI + specialization: coding + + ########################################################################## + # Meta (open-weight) + ########################################################################## + META: + description: Open-weight models developed by Meta (the Llama family) + annotations: + node_type: developer + homepage: https://www.llama.com + license: Llama Community License + see_also: + - https://www.llama.com/models/ + + LLAMA_4: + description: Llama 4 - Meta's open-weight mixture-of-experts model family + is_a: META + annotations: + node_type: family + + llama-4-maverick: + description: >- + Llama 4 Maverick - general-purpose open-weight chat model; highest MMLU + among Llama 4 variants. + is_a: LLAMA_4 + title: Llama 4 Maverick + aliases: + - Llama 4 Maverick + annotations: + node_type: model + api_model_id: llama-4-maverick + developer: Meta + + llama-4-scout: + description: >- + Llama 4 Scout - open-weight model supporting very long context + (up to 10M tokens). + is_a: LLAMA_4 + title: Llama 4 Scout + aliases: + - Llama 4 Scout + annotations: + node_type: model + api_model_id: llama-4-scout + developer: Meta + context_window: 10000000 + + ########################################################################## + # DeepSeek (open-weight) + ########################################################################## + DEEPSEEK: + description: Open-weight models developed by DeepSeek + annotations: + node_type: developer + homepage: https://www.deepseek.com + see_also: + - https://api-docs.deepseek.com/ + + DEEPSEEK_MODELS: + description: DeepSeek - open-weight mixture-of-experts chat and reasoning models + is_a: DEEPSEEK + annotations: + node_type: family + + deepseek-chat: + description: >- + DeepSeek-V3 (served as deepseek-chat) - open-weight MoE model activating + ~37B of 671B parameters per token. + is_a: DEEPSEEK_MODELS + title: DeepSeek-V3 + aliases: + - DeepSeek-V3 + - DeepSeek V3 + annotations: + node_type: model + api_model_id: deepseek-chat + developer: DeepSeek + + deepseek-reasoner: + description: >- + DeepSeek-R1 (served as deepseek-reasoner) - open-weight reasoning model. + is_a: DEEPSEEK_MODELS + title: DeepSeek-R1 + aliases: + - DeepSeek-R1 + - DeepSeek R1 + annotations: + node_type: model + api_model_id: deepseek-reasoner + developer: DeepSeek + + ########################################################################## + # Mistral AI (open-weight) + ########################################################################## + MISTRAL: + description: Models developed by Mistral AI + annotations: + node_type: developer + homepage: https://mistral.ai + see_also: + - https://docs.mistral.ai/getting-started/models/models_overview/ + + MISTRAL_LARGE: + description: Mistral Large - Mistral AI's flagship open-weight model family + is_a: MISTRAL + annotations: + node_type: family + + mistral-large-3: + description: >- + Mistral Large 3 - state-of-the-art open-weight MoE model (41B active / + 675B total parameters), 256K context, released under Apache 2.0. + is_a: MISTRAL_LARGE + title: Mistral Large 3 + aliases: + - Mistral Large 3 + annotations: + node_type: model + api_model_id: mistral-large-3 + developer: Mistral AI + release_date: "2025-12" + context_window: 256000 + license: Apache-2.0 + + ########################################################################## + # Moonshot AI (open-weight) + ########################################################################## + MOONSHOT: + description: Models developed by Moonshot AI (the Kimi family) + annotations: + node_type: developer + homepage: https://www.moonshot.ai + see_also: + - https://platform.moonshot.ai/docs + + KIMI: + description: Kimi - Moonshot AI's open-weight mixture-of-experts model family + is_a: MOONSHOT + annotations: + node_type: family + + kimi-k2.6: + description: >- + Kimi K2.6 - frontier open-weight MoE coding model + (32B active / 1T total parameters). + is_a: KIMI + title: Kimi K2.6 + aliases: + - Kimi K2.6 + annotations: + node_type: model + api_model_id: kimi-k2.6 + developer: Moonshot AI diff --git a/src/valuesets/schema/valuesets.yaml b/src/valuesets/schema/valuesets.yaml index 0859ea47..8dbb3337 100644 --- a/src/valuesets/schema/valuesets.yaml +++ b/src/valuesets/schema/valuesets.yaml @@ -58,6 +58,7 @@ imports: - data/data_use - computing/mime_types - computing/quantum +- computing/frontier_ai_models - computing/microelectronics - computing/data_centers - physics/states_of_matter