Anthropic model product
Claude Opus 4.6
6 is Anthropic's most intelligent model, improving on its predecessor's coding skills with more careful planning, longer agentic task sustenance, more reliable operation in larger codebases, and better code review and debugging skills.
Updated Aug 12, 2026. Default version: Claude Opus 4.6
LLMBoard score79.6Claude Opus 4.6
Coverage80%26 benchmark families
Context window1MTokens
Official input price$5Anthropic API
Claude Opus 4.6 Specifications
Technical details for the model's default version.
- Version
- Claude Opus 4.6
- Released
- Feb 5, 2026
- Knowledge cutoff
- May 31, 2025
- Parameters
- N/A
- Context window
- 1M
- Max output
- 128K
- Inputs
- image, pdf, text
- Outputs
- text
- Open weights
- No
- License
- Proprietary
Claude Opus 4.6 Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Claude Opus 4.6 Benchmark Results
Benchmark scores for Claude Opus 4.6.
| Graphwalks parents >128k | 95.4% | 01 | 7 | 100.0% | C | |
| OpenRCA | 34.9% | 01 | 1 | 100.0% | C | |
| Tau2 Retail | 91.9% | 01 | 26 | 100.0% | C | |
| Tau2 Telecom | 99.3% | 01 | 35 | 100.0% | C | |
| Vending-Bench 2 | 8,017.59 usd | 01 | 4 | 100.0% | C | |
| FigQA | 78.3% | 02 | 3 | 50.0% | C | |
| DeepSearchQA | 91.3% | 03 | 9 | 75.0% | C | |
| Finance Agent | 60.7% | 03 | 8 | 71.4% | C | |
| MRCR v2 (8-needle) | 76.0% | 03 | 21 | 90.0% | C | |
| OSWorld | 72.7% | 03 | 20 | 89.5% | C | |
| ARC-AGI v2 | 68.8% | 05 | 16 | 73.3% | C | |
| SWE-bench Multilingual | 77.8% | 05 | 34 | 87.9% | C | |
| AIME 2025 | 99.8% | 06 | 114 | 95.6% | C | |
| CyberGym | 73.8% | 06 | 11 | 50.0% | C | |
| Graphwalks BFS >128k | 61.5% | 06 | 11 | 50.0% | C | |
| Legal Agent Benchmark | 4.2% | 06 | 13 | 58.3% | B | |
| MMMLU | 91.1% | 06 | 49 | 89.6% | C | |
| SWE-Bench Verified | 80.8% | 07 | 105 | 94.2% | C | |
| FrontierSWE | 56.0% | 08 | 15 | 50.0% | B | |
| LiveBench | 76.3% | 11 | 38 | 73.0% | B | |
| Humanity's Last Exam | 53.1% | 14 | 93 | 85.9% | C | |
| BrowseComp | 84.0% | 15 | 58 | 75.4% | C | |
| GPQA | 91.3% | 17 | 234 | 93.1% | C | |
| Terminal-Bench 2.0 | 65.4% | 17 | 49 | 66.7% | C | |
| MMMU-Pro | 77.3% | 23 | 66 | 66.2% | C | |
| MCP Atlas | 62.7% | 24 | 31 | 23.3% | C | |
| CharXiv-R | 77.4% | 28 | 48 | 42.5% | C |
Claude Opus 4.6 Arena Results
Preference and agent-evaluation results for the default version.
| search | overall | 01 | 1252.8 | 112,201 | N/A | |
| document | overall | 02 | 1509.9 | 37,271 | N/A | |
| search factuality | overall | 02 | 1227.3 | 112,157 | N/A | |
| text factuality | overall | 02 | 1492.9 | 71,652 | N/A | |
| text style control | overall | 02 | 1505.0 | 71,992 | N/A | |
| document | overall | 03 | 1505.9 | 24,973 | N/A | |
| search style control | overall | 03 | 1222.7 | 112,201 | N/A | |
| text | overall | 03 | 1503.0 | 71,992 | N/A | |
| text factuality | overall | 03 | 1487.5 | 75,618 | N/A | |
| document style control | overall | 04 | 1494.7 | 37,271 | N/A | |
| text | overall | 04 | 1497.4 | 75,983 | N/A | |
| vision style control | overall | 04 | 1300.1 | 20,310 | N/A | |
| document style control | overall | 05 | 1493.0 | 24,973 | N/A | |
| text style control | overall | 05 | 1497.6 | 75,983 | N/A | |
| vision | overall | 05 | 1315.8 | 20,310 | N/A | |
| vision | overall | 07 | 1311.0 | 24,609 | N/A | |
| vision style control | overall | 09 | 1292.9 | 24,609 | N/A | |
| agent bash recovery steps | overall | 10 | 0.1 | N/A | 24.2K | |
| agent steerability | overall | 11 | 0.1 | N/A | 34.3K | |
| agent | overall | 13 | 0.1 | N/A | 1.2M | |
| agent tool hallucination | overall | 13 | 0.0 | N/A | 1.1M | |
| webdev | overall | 13 | 1545.0 | 15,936 | N/A | |
| agent praise complaint | overall | 14 | 0.1 | N/A | 9.1K | |
| webdev | overall | 16 | 1537.6 | 17,166 | N/A | |
| agent task outcome explicit | overall | 17 | 0.0 | N/A | 26.9K |
Claude Opus 4.6 Runtime Performance
Provider-specific output speed and catalog latency for Claude Opus 4.6. Runtime does not affect the capability score.
| Anthropic | 3.488 tok/s | 7.164 s | 1M | 128K |
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Claude Opus 4.6 Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- $5 input, $25 output per 1M
- Official provider
- Anthropic
- Lowest third-party
- From $4.3 input, $21 output per 1M via Poe
- Tracked offerings
- 37
| Poe | anthropic/claude-opus-4.6 | global | $4.3 | $21 | 983K | |
| Impossibl | anthropic/claude-opus-4-6 | global | $5 | $25 | 1M | |
| NEAR AI Cloud | anthropic/claude-opus-4-6 | global | $5 | $25 | 200K | |
| LLM Gateway | claude-opus-4-6 | global | $5 | $25 | 1M | |
| OpenCode Zen | claude-opus-4-6 | global | $5 | $25 | 1M | |
| GitHub Copilot | claude-opus-4.6 | global | $5 | $25 | 200K | |
| 302.AI | claude-opus-4-6 | global | $5 | $25 | 1M | |
| Neon | claude-opus-4-6 | global | $5 | $25 | 1M | |
| FrogBot | claude-opus-4-6 | global | $5 | $25 | 200K | |
| Pioneer | claude-opus-4-6 | global | $5 | $25 | 1M | |
| ZenMux | anthropic/claude-opus-4.6 | global | $5 | $25 | 1M | |
| Abacus | claude-opus-4-6 | global | $5 | $25 | 1M | |
| Azure Cognitive Services | claude-opus-4-6 | global | $5 | $25 | 1M | |
| Auriko | claude-opus-4-6 | global | $5 | $25 | 1M | |
| Azure | claude-opus-4-6 | global | $5 | $25 | 1M | |
| Vertex (Anthropic) | claude-opus-4-6@default | global | $5 | $25 | 1M | |
| AIHubMix | claude-opus-4-6 | global | $5 | $25 | 1M | |
| Ofox | anthropic/claude-opus-4.6 | global | $5 | $25 | 1M | |
| OrcaRouter | anthropic/claude-opus-4.6 | global | $5 | $25 | 1M | |
| Vertex | claude-opus-4-6@default | global | $5 | $25 | 1M | |
| GMI Cloud | anthropic/claude-opus-4.6 | global | $5 | $25 | 409.6K | |
| FreeModel | claude-opus-4-6 | global | $5 | $25 | 1M | |
| Amazon Bedrock | anthropic.claude-opus-4-6-v1 | global | $5 | $25 | 1M | |
| Requesty | claude-opus-4-6 | global | $5 | $25 | 1M | |
| Kilo Gateway | anthropic/claude-opus-4.6 | global | $5 | $25 | 1M | |
| Merge Gateway | anthropic/claude-opus-4-6 | global | $5 | $25 | 1M | |
| Vercel AI Gateway | anthropic/claude-opus-4.6 | global | $5 | $25 | 1M | |
| OpenRouter | anthropic/claude-opus-4.6 | global | $5 | $25 | 1M | |
| Anthropic | claude-opus-4-6 | global | $5 | $25 | 1M | |
| Cloudflare AI Gateway | anthropic/claude-opus-4-6 | global | $5 | $25 | 1M | |
| Jiekou.AI | claude-opus-4-6 | global | $5 | $25 | 1M | |
| Perplexity Agent | anthropic/claude-opus-4-6 | global | $5 | $25 | 200K | |
| Databricks | databricks-claude-opus-4-6 | global | $5 | $25 | 1M | |
| Cortecs | claude-opus4-6 | global | $5.31 | $26.56 | 1M | |
| Venice AI | claude-opus-4-6 | global | $6 | $30 | 1M |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Claude Opus 4.6 Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
| Claude Opus 4.6 | 79.6 | N/A | 1M | 128K | No | Proprietary |
What is Claude Opus 4.6?
Key information about Claude Opus 4.6 and its available data.
6 is Anthropic's most intelligent model, improving on its predecessor's coding skills with more careful planning, longer agentic task sustenance, more reliable operation in larger codebases, and better code review and debugging skills.
First Opus-class model with 1M token context window (beta), 128K output tokens, and adaptive thinking. Features effort controls (low/medium/high/max) and context compaction for long-running tasks. 0, Humanity's Last Exam, GDPval-AA, and BrowseComp. Pricing: $5/$25 per million tokens (input/output).
Data as of 2026-08-11.
Claude Opus 4.6 vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to Claude Opus 4.6
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
AN
Claude Sonnet 5
Anthropic
AN
Claude Opus 4.7
Anthropic
AN
Claude Sonnet 4.6
Anthropic
AN
Claude Opus 4.8
Anthropic
AN
Claude Opus 4.5
Anthropic
AN
Claude Mythos
Anthropic
FAQ
Common questions about Claude Opus 4.6.
When was Claude Opus 4.6 released?
Claude Opus 4.6's default version was released on Feb 5, 2026.
How much does Claude Opus 4.6 cost?
Claude Opus 4.6's official API price is $5 per million input tokens and $25 per million output tokens via Anthropic. The lowest tracked third-party offer starts at $4.3 input and $21 output via Poe.
Who created Claude Opus 4.6?
Claude Opus 4.6 was created by Anthropic.
What is the context window for Claude Opus 4.6?
The default version has a 1M token context window.
Is Claude Opus 4.6 open weight?
No. The default version is not marked as having publicly available weights.
How many API providers offer Claude Opus 4.6?
37 provider offerings are linked to the default version.
What models should I compare Claude Opus 4.6 with?
Nearby ranked alternatives include GPT-5.5, Muse Spark 1.2, Grok 4.5.