last verified: 2026-04-09 | sources: claude.ai plans, cursor pricing, codex pricing, gemini subscriptions, antigravity
| tool | free | entry | mid-tier | heavy use | teams |
|---|---|---|---|---|---|
| claude code | limited | $20/mo (Pro) | $100/mo (Max 5x) | $200/mo (Max 20x) | enterprise (API) |
| cursor | limited (2-week trial) | $20/mo (Pro) | $60/mo (Pro+) | $200/mo (Ultra) | $40/user/mo |
| codex | limited (temporary) | $20/mo (ChatGPT Plus) | -- | $200/mo (ChatGPT Pro) | $30/user/mo (Business) |
| gemini | generous (60 req/min) | $19.99/mo (AI Pro) | -- | $249.99/mo (AI Ultra) | Vertex AI / GCP pricing |
| antigravity | free (public preview) | $0 (preview) | -- | custom (enterprise) | custom |
| tool | what you get | verdict |
|---|---|---|
| claude code | limited requests, rate-throttled | evaluation only |
| cursor | 2-week pro trial, 2000 completions, 50 slow requests | trial, not sustained use |
| codex | limited codex access (temporary offer) | trial |
| gemini | 60 req/min, 1000 req/day, no credit card | real daily usage |
| antigravity | free during public preview, weekly quotas, multi-model access | real daily usage (best free tier) |
winner: antigravity. free access to gemini 3.1 pro, claude opus/sonnet, and GPT-OSS 120B during public preview. gemini is second with 1000 requests/day on its own models.
| tool | plan | what you get |
|---|---|---|
| claude code | Pro ($20) | full claude code access, all models, rate-limited |
| cursor | Pro ($20) | unlimited completions, 500 fast premium requests, credit pool |
| codex | ChatGPT Plus ($20) | codex access, GPT-4o, web browsing, plugins |
| gemini | AI Pro ($19.99) | higher limits, workspace AI, 2TB storage |
| antigravity | $0 (preview) | free access to frontier models, weekly quotas |
winner: depends on workflow. cursor gives you tab completion + chat + agent. claude code gives you terminal agent + extensibility. codex gives you the broadest openai product access. gemini gives you cloud storage and workspace integration. antigravity is still free (preview).
| tool | plan | what you get |
|---|---|---|
| claude code | Max 5x ($100) / Max 20x ($200) | 5x or 20x Pro rate limits, max priority |
| cursor | Pro+ ($60) / Ultra ($200) | larger credit pools, higher rate limits |
| codex | ChatGPT Pro ($200) | 2x codex limits, o3-pro, GPT-5, all openai models |
| gemini | AI Ultra ($249.99) | highest-tier models, all google AI features |
| antigravity | $0 (preview) / custom (enterprise) | free access continues, enterprise pricing TBD |
winner for individuals: claude code Max 5x at $100/mo hits a sweet spot -- 5x the throughput of Pro for real power users. no other tool has an equivalent mid-tier. cursor's Pro+ at $60/mo is the closest.
subscription pricing tells you the monthly bill. but effective cost depends on how much you use the tool.
session cost varies widely based on model, duration, and cache efficiency -- use /lore to see your actual per-session costs. on the $20/mo Pro plan, even moderate usage typically exceeds the subscription cost in raw API value. on Max 20x at $200/mo, anthropic is subsidizing heavy users -- a full day of opus sessions can exceed $200 in raw API costs.
the lore plugin in this repo tracks actual per-session costs, so you can measure this precisely.
cursor's credit system makes this harder to calculate. a Pro user gets $20 worth of credits. using claude opus burns credits faster than cursor-small. heavy cursor users report running out of credits mid-month and needing to upgrade or switch to slower models.
ChatGPT Plus at $20/mo includes codex with rate limits. heavy users report hitting limits during sustained coding sessions and needing to wait or upgrade to Pro at $200/mo. the jump from $20 to $200 with no middle tier is steep.
the free tier is generous enough that many developers never need to pay. 1000 requests/day is hard to exhaust in normal use. if you do need more, API key access with pay-per-token is available as an alternative to subscription.
currently free during public preview. weekly quotas limit heavy usage but most developers won't hit them. when google announces post-preview pricing, session economics will depend on the pricing model they choose. google AI Pro/Ultra subscribers get priority access and higher quotas.
hidden costs and gotchas
- opus is 5x the cost of sonnet. switching to opus mid-session for "just one question" can spike costs
- long sessions (30+ turns) accumulate context that increases per-turn cost even with caching
- subagents multiply billing -- each runs its own context window
- credit system makes costs unpredictable -- depends on model mix
- running out of credits mid-month forces model downgrades or upgrade
- annual billing saves 20% but locks you in
- no middle tier between $20 and $200 -- the jump is 10x
- API key usage has separate pricing from ChatGPT subscription
- cloud agent execution time counts against limits
- AI Ultra at $249.99/mo is the most expensive individual tier across all tools
- google ecosystem lock-in -- Vertex AI auth, GCP integration create switching costs
- free tier limits may tighten as adoption grows
- public preview pricing will change -- google hasn't committed to free forever
- weekly quotas throttle heavy users during preview
- post-preview pricing model is unknown -- could be subscription, credits, or usage-based
openai codex continues alpha releases (0.115.0-alpha.15-17) but no stable version bumps. gemini-cli active development with oauth2 and headless fixes. no pricing changes detected across codex, gemini, or antigravity. claude code remains the dominant terminal agent for cost-conscious teams.
use antigravity (free preview, includes frontier models) as your primary tool. supplement with gemini free tier for terminal workflows.
if terminal-first: claude code Pro ($20) if IDE-first: cursor Pro ($20) or antigravity (still free) if you want everything openai: ChatGPT Plus ($20)
power user: claude code Max 5x ($100) -- best value at this tier IDE power user: cursor Pro+ ($60)
claude code Max 20x: best for developers who live in claude code all day cursor Ultra: best for developers who live in cursor all day ChatGPT Pro: best for openai ecosystem access beyond just coding
if you're spending this much, you should be tracking per-session costs. use the lore plugin to measure actual usage and optimize model selection.
individual comparison docs: codex | cursor | gemini | antigravity