tokeven16 models across 3 tiers
Claude API pricing in 2026
Anthropic has the Claude model family spanning three tiers: flagship (Opus 4-8, Fable 5), mid-tier (Sonnet 4-6, Sonnet 5), and fast (Haiku 4-5). Claude models score well on code generation and reasoning benchmarks. All Claude models have a 90% cache read discount and a 1.25x cache write surcharge.
Models
16
Cheapest input
$0.10/MTok
Cache discount
90% off
Max context
1M
All Claude models
| Model | Tier | Input $/MTok | Output $/MTok | Context | Best task | Released |
|---|---|---|---|---|---|---|
| Claude Haiku 5.5 | Fast / Budget | $0.10 | $0.50 | 1M | Not benchmarked | Oct 2026 |
| Claude Haiku 4.5 | Fast / Budget | $1.00 | $5.00 | 200K | Classification (90/100) | Oct 2025 |
| Claude Sonnet 5 | Mid-tier | $2.00 | $10.00 | 1M | Code Review (96/100) | Jun 2026 |
| Claude Sonnet 5.5 | Mid-tier | $2.00 | $10.00 | 1M | Not benchmarked | Sep 2026 |
| Claude Sonnet 4.6 | Mid-tier | $3.00 | $15.00 | 1M | Code Review (95/100) | May 2026 |
| Claude Sonnet 4.5 | Mid-tier | $3.00 | $15.00 | 200K | Not benchmarked | — |
| Claude Opus 5.5 | Flagship | $4.00 | $20.00 | 1M | Reasoning (92/100) | Sep 2026 |
| Claude Opus 4.8 | Flagship | $5.00 | $25.00 | 1M | Code Review (94/100) | Apr 2026 |
| Claude Opus 5 | Flagship | $5.00 | $25.00 | 1M | Reasoning (91/100) | Jul 2026 |
| Claude Opus 4.7 | Flagship | $5.00 | $25.00 | 1M | Reasoning (87/100) | — |
| Claude Opus 4.6 | Flagship | $5.00 | $25.00 | 1M | Reasoning (89/100) | — |
| Claude Opus 4.5 | Flagship | $5.00 | $25.00 | 200K | Not benchmarked | — |
| Claude Fable 5.1 | Flagship | $10.00 | $50.00 | 1M | Reasoning (92/100) | Aug 2026 |
| Claude Fable 5 | Flagship | $10.00 | $50.00 | 1M | Code Review (95/100) | Jun 2026 |
| Claude Mythos 5.1 | Flagship | $10.00 | $50.00 | 1M | Not benchmarked | — |
| Claude Mythos 5 | Flagship | $10.00 | $50.00 | 1M | Not benchmarked | — |
Claude Haiku 5.5: Added by the model-onboarding job on 2026-10-08. Prices confirmed by the provider pricing page and the LiteLLM cross-check. Release date from the provider model list. Not yet benchmarked, so never recommended.
Claude Sonnet 5: Launched at $2/$10 as introductory pricing through Aug 31, 2026. Anthropic cancelled the scheduled increase to $3/$15, so this is now the standard price.
Claude Sonnet 5.5: Added by the model-onboarding job on 2026-09-29. Prices confirmed by the provider pricing page and the LiteLLM cross-check. Release date from the provider model list. Not yet benchmarked, so never recommended.
Claude Sonnet 4.5: Added 2026-09-27 after the backfill hold. Prices from the provider pricing page; 200K window (no long-context tier) from the provider context-windows page. Not yet benchmarked, so never recommended.
Claude Opus 5.5: Added by the model-onboarding job on 2026-09-24. Prices confirmed by the provider pricing page and the LiteLLM cross-check. Release date from the provider model list. Benchmarked from the LiveBench 2026-06-25 release (2026-W26) in the first rankings sync.
Claude Opus 4.7: Backfilled on 2026-09-23. Prices confirmed by the provider pricing page and the LiteLLM cross-check. Release date not published in a machine-readable source. Benchmarked from the LiveBench 2026-06-25 release (2026-W26) in the first rankings sync.
Claude Opus 4.6: Backfilled on 2026-09-23. Prices confirmed by the provider pricing page and the LiteLLM cross-check. Release date not published in a machine-readable source. Benchmarked from the LiveBench 2026-06-25 release (2026-W26) in the first rankings sync.
Claude Opus 4.5: Backfilled on 2026-09-23. Prices confirmed by the provider pricing page and the LiteLLM cross-check. Release date not published in a machine-readable source. Not yet benchmarked, so never recommended.
Claude Mythos 5.1: Backfilled on 2026-09-23. Prices confirmed by the provider pricing page and the LiteLLM cross-check. Release date not published in a machine-readable source. Not yet benchmarked, so never recommended.
Claude Mythos 5: Backfilled on 2026-09-23. Prices confirmed by the provider pricing page and the LiteLLM cross-check. Release date not published in a machine-readable source. Not yet benchmarked, so never recommended.
Claude cost by use case
Estimated cost per API call across common workloads, using default token counts for each use case.
Claude quality benchmarks by task
Confidence scores (0-100) across all task types. Higher is better.
Confidence scores are a third-party benchmark snapshot from week 26 of 2026. This snapshot is 17 weeks old, so treat the scores as directional and validate against your own evals.
| Task | 5.5 | 4.5 | 5 | 5.5 | 4.6 | 4.5 | 5.5 | 4.8 | 5 | 4.7 | 4.6 | 4.5 | 5.1 | 5 | 5.1 | 5 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Code Generation | - | 72/100 | 68/100 | - | 57/100 | - | 79/100 | 63/100 | 72/100 | 63/100 | 61/100 | - | 74/100 | 72/100 | - | - |
| Code Review | - | 70/100 | 96/100 | - | 95/100 | - | - | 94/100 | - | - | - | - | - | 95/100 | - | - |
| Summarization | - | 82/100 | 64/100 | - | 61/100 | - | 66/100 | 69/100 | 65/100 | 66/100 | 62/100 | - | 71/100 | 77/100 | - | - |
| Q&A | - | 86/100 | 89/100 | - | 88/100 | - | - | 85/100 | - | - | - | - | - | 88/100 | - | - |
| Extraction | - | 88/100 | 70/100 | - | 71/100 | - | 75/100 | 73/100 | 73/100 | 73/100 | 73/100 | - | 75/100 | 75/100 | - | - |
| Reasoning | - | 65/100 | 89/100 | - | 85/100 | - | 92/100 | 89/100 | 91/100 | 87/100 | 89/100 | - | 92/100 | 90/100 | - | - |
| Classification | - | 90/100 | 86/100 | - | 84/100 | - | - | 80/100 | - | - | - | - | - | 82/100 | - | - |
| Creative Writing | - | 60/100 | 62/100 | - | 69/100 | - | 71/100 | 81/100 | 61/100 | 69/100 | 66/100 | - | 78/100 | 72/100 | - | - |
Recent Claude pricing changes
Claude Haiku 5.5 input launched
$0.10/MTok at launch
Claude Sonnet 5.5 input launched
$2.00/MTok at launch
Claude Opus 5.5 input launched
$4.00/MTok at launch
Claude Sonnet 5 input decreased
$3.00 to $2.00/MTok
Claude Sonnet 5 output decreased
$15.00 to $10.00/MTok
Claude Sonnet 5 input launched
$3.00/MTok at launch
Compare Claude with other providers
Head-to-head comparisons
Track your Claude spend
See exactly how much you spend on Claude models and where you can save. Two-line wrapper install, no API keys shared.