Claude (Anthropic) Pricing
All plans and pricing as of 2026-09-28
Free
- ✓Limited messages/day
- ✓Claude Sonnet 5 (default as of 2026-06-30); Sonnet 5.5 shipped 2026-09-28 at the same API price
- ✓Basic features
Pro
- ✓5x more usage than Free
- ✓Claude Opus 5.5 (2026-09-22) -- the new leading Claude model, Fable 5.1-level on most work; Opus 5 remains
- ✓Sonnet 5 / Sonnet 5.5 (2026-09-28) for everyday work
- ✓Five-hour usage limits raised 2026-09-22 plus a saveable rate-limit reset
- ✓Effort control + extended thinking
- ✓Priority access
Max (5x)
- ✓5x Pro usage
- ✓Priority queue
- ✓Opus 5.5 (2026-09-22) succeeds Opus 5 as the leading model; five-hour limits raised with a saveable rate-limit reset
- ✓Full effort control + fast mode
Max (20x)
- ✓20x Pro usage
- ✓Highest priority
- ✓All generally-available models
- ✓Best for power users and agents
API (Opus 5.5, launched 2026-09-22)
- ✓Fable 5.1-level on most work per Anthropic; 40% cheaper than Opus 5 on typical workloads at default settings
- ✓Fast mode $8 / $40 per 1M for up to 2.5x speed (Claude Code + first-party API only)
- ✓1M context, Jun 2026 cutoff, Batch $2 / $10; `claude-opus-5-5` on Bedrock, Vertex AI and Microsoft Foundry day one
- ✓Ships with Fable-class cyber/bio safeguards (cyber tasks fall back to Opus 4.8) and thinking cannot be turned off
API (Sonnet 5.5, launched 2026-09-28)
- ✓Up to 30% cheaper per task than Sonnet 5 through fewer tokens, output 30%+ faster
- ✓Terminal-Bench 4.0 70.6% (Sonnet 5: 10.3%); CursorBench 4.0 55.5%; GDPval-AA 1844 vs Opus 5.5's 1846 (vendor numbers)
- ✓1M context, Jun 2026 cutoff, Batch $1 / $5; `claude-sonnet-5-5` on all platforms
- ✓First Sonnet with cyber safeguards and reasoning-extraction classifiers; thinking-off users must adopt `between_tools`
API (Opus 5)
- ✓Launched 2026-07-24 as `claude-opus-5`; same rate as Opus 4.8, half of Fable 5's $10/$50
- ✓Default model on Claude Max, strongest model on Claude Pro; also on Claude Code and Claude Cowork
- ✓Anthropic's stated positioning: within 0.5% of Fable 5's peak CursorBench 3.2 score at half the cost per task
- ✓Available via the Claude API, Google Cloud, and Bedrock (anthropic.claude-opus-5)
API (Opus 4.8)
- ✓Unchanged from Opus 4.7 pricing
- ✓1M context window
- ✓Fast mode at $10 / $50 per 1M (2.5x speed, 3x cheaper than prior fast mode)
- ✓Tool use, MCP, high-res vision; Bedrock, Vertex AI, Foundry
API (Sonnet 5)
- ✓Launched 2026-06-30 as the 'most agentic Sonnet yet' -- approaches Opus 4.8 quality at lower cost
- ✓Default model on Free + Pro; also on Max/Team/Enterprise, Claude Code, and the API (claude-sonnet-5)
- ✓Updated tokenizer (input maps to ~1.0-1.35x more tokens depending on content type)
- ✓OSWorld-Verified 78.5%; scored 0.0% on cyber-exploit-development evals (safety)
API (Fable 5)
- ✓First publicly available Mythos-class model (launched 2026-06-09; suspended 6/12 by US-gov order; RESTORED globally 2026-07-01 after controls lifted 6/30)
- ✓PERMANENT plan split from 2026-07-20 (announced 7/18): Max + Team Premium keep Fable 5 included at 50% of usage limits; Pro + Team Standard move to pay-as-you-go credits with a one-time $100 grant
- ✓API rate $10/$50 per 1M; tiered credit bundles discount up to 30%
- ✓Auto-fallback on cyber/bio/chem-flagged requests; **as of 2026-08-07 the biology classifier was retuned and biology fallbacks dropped ~85%**, and the stated fallback target is now **Opus 5** (dual-use virology/toxicology/molecular design still fall back); mandatory 30-day retention on Mythos-class traffic (not used for training)
Is Claude (Anthropic) Worth the Price?
Value Score: 8/10
Overall Score: 8.5/10 · Writers, analysts, developers, and anyone who values quality of output over quantity of features. If you care about how good the actual text is, Claude is the best.
Claude is the LLM you pick when quality matters more than features, and the Claude 5.5 family (Opus 5.5 on 2026-09-22, Sonnet 5.5 on 2026-09-28) is the strongest version of that argument yet. Opus 5.5 does Fable 5.1-level work at $4/$20 with $0.20 cache reads -- 40% cheaper than Opus 5 on typical workloads by Anthropic's measurement -- and leads its own table on Terminal-Bench 4.0, FrontierCode, CursorBench and GDPval-AA, though Anthropic is unusually candid that the real-world gap to Fable 5.1 is narrower than the numbers. Sonnet 5.5 is the bigger practical shift for most people: the $2/$10 model now posts 70.6% on Terminal-Bench 4.0 where Sonnet 5 managed 10.3%, and it does it with fewer tokens and faster. The catch is the safeguard regime that now reaches down to Opus and, for cyber, to Sonnet: most security work is re-routed to older models unless you are verified, and thinking cannot be disabled on Opus 5.5. Fable 5.1 stays the top model at $10/$50 for the hardest long-horizon work; for everything else the 5.5 pair is the better buy. Scores held at 8.5 pending third-party verification of the vendor table.
The Tier List Tuesday
Weekly newsletter: tier movers, new entrants, and the VS of the week. Built from our daily AI-tool sweeps. No spam, unsubscribe anytime.
How Claude (Anthropic) Pricing Compares
| Tool | Free Tier | Starting Price | Value Score | Overall |
|---|---|---|---|---|
| Claude (Anthropic)(this tool) | Yes | $0 | 8/10 | 8.5 |
| Muse Spark (Meta) | Yes | $0 | 10/10 | 8.8 |
| Gemini (Google) | Yes | $0 | 9/10 | 8.3 |
| MiMo (Xiaomi) | Yes | $0 | 9/10 | 8.3 |
| Hunyuan 3 (Tencent Hy3) | Yes | $0 | 9.5/10 | 8.1 |
| Grok | Yes | $0 | 7.5/10 | 7.5 |