BlogLLM Benchmark Scores as Published, Side by Side
Claude Sonnet 4.5: Benchmark Scores as Published
Every result the catalog holds for Claude Sonnet 4.5: 87 tests in 14 areas, from 15 pages. 21 printed by Anthropic itself, the rest by the company named beside each number. Each number is as printed; nothing is ranked or rescored.
Developer AnthropicReleased 2025-09-29Context window 200KMax output 64KInput Text, imageOutput TextWeights ClosedData version 2026103001, as of 2026-10-30
Knowledge Work
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GDPval-AA (Elo) | 1277 | Anthropic, Claude Opus 4.6checked 2026-10-15; printed on 1 more page | - |
| Vals Finance Agent v1.1 | 54.5% | Anthropic, Claude Sonnet 4.6checked 2026-10-15 | - |
| Vals Finance Agent (version not stated) | 55.3% | Anthropic, Claude Sonnet 4.5checked 2026-10-15; printed on 1 more page | Effort: extended thinking (up to 64K) |
Agentic Coding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| SWE-bench Verified | 78.2% | Anthropic, Claude Sonnet 4.5checked 2026-10-15; printed on 9 more pages | Effort: thinking budget 200K; Context: 1M |
| 82% | Anthropic, Claude Sonnet 4.5checked 2026-10-15 | Effort: thinking budget 200K; Attempt budget: parallel test-time compute: multiple samples, the best selected by an internal scoring model | |
| 77.2% | Anthropic, Claude Sonnet 4.5checked 2026-10-15 | Effort: thinking budget 200K | |
| SWE-bench Multilingual | 68% | Moonshot AI, Kimi K2 Thinking release pagechecked 2026-10-15; printed on 1 more page | Effort: thinking; Harness: in-house harness derived from SWE-agent; Tools: enabled; bash and edit tools |
| Terminal-Bench 2.0 | 50% | Anthropic, Claude Opus 4.5checked 2026-10-15; printed on 4 more pages | - |
| Terminal-Bench (version not stated) | 50% | Anthropic, Claude Sonnet 4.5checked 2026-10-15; printed on 2 more pages | - |
| Terminal-Bench Hard | 33.3% | Z.ai, GLM-4.7 model cardchecked 2026-10-15 | - |
| Multi-SWE-bench | 44.3% | Moonshot AI, Kimi K2 Thinking release pagechecked 2026-10-15 | Effort: thinking; Harness: in-house harness derived from SWE-agent; Tools: enabled; bash and edit tools |
Science and Math
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPQA Diamond | 83.4% | Anthropic, Claude Sonnet 4.5checked 2026-10-15; printed on 7 more pages | - |
| AIME 2025 | 87% | Anthropic, Claude Sonnet 4.5checked 2026-10-15; printed on 6 more pages | Tools: none |
| 100% | Anthropic, Claude Sonnet 4.5checked 2026-10-15 | Effort: 64K reasoning tokens; Tools: Python | |
| SciCode | 45% | MiniMax, MiniMax-M2.5 model cardchecked 2026-10-15; printed on 1 more page | - |
| HMMT February 2025 | 79.2% | DeepSeek, DeepSeek-V3.2 model cardchecked 2026-10-15; printed on 2 more pages | - |
| IMO-AnswerBench | 65.9% | Moonshot AI, Kimi K2 Thinking release pagechecked 2026-10-15; printed on 1 more page | Effort: thinking; Tools: none |
| HMMT November 2025 | 81.7% | Z.ai, GLM-4.7 model cardchecked 2026-10-15 | - |
| BeyondAIME | 59.8% | Mistral AI, Mistral Medium 3.5 model cardchecked 2026-10-15 | Effort: max |
Long Context
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| AA-LCR | 66% | MiniMax, MiniMax-M2.5 model cardchecked 2026-10-15 | - |
Computer Use
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| OSWorld-Verified | 61.4% | Anthropic, Claude Sonnet 4.5checked 2026-10-15; printed on 5 more pages | - |
| ScreenSpot-Pro | 36.2% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
Multimodal Understanding
General Knowledge and Reasoning
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Humanity's Last Exam (no tools) | 17.7% | Anthropic, Claude Opus 4.6checked 2026-10-15; printed on 5 more pages | Effort: max; Tools: none |
| MMLU-Pro | 87.5% | Moonshot AI, Kimi K2 Thinking release pagechecked 2026-10-15; printed on 1 more page | Effort: thinking; Tools: none |
| Humanity's Last Exam (with tools) | 33.6% | Anthropic, Claude Opus 4.6checked 2026-10-15; printed on 3 more pages | Effort: max; Tools: enabled; exact tools in source |
| MMLU-Redux | 95.6% | Moonshot AI, Kimi K2 Thinking release pagechecked 2026-10-15 | Effort: thinking; Tools: none |
| ARC-AGI-2 | 13.6% | Anthropic, Claude Opus 4.5checked 2026-10-15; printed on 2 more pages | - |
Code Generation
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| LiveCodeBench v6 | 64% | Moonshot AI, Kimi K2 Thinking release pagechecked 2026-10-15; printed on 1 more page | Effort: thinking; Tools: none |
| Codeforces rating | 1480 | DeepSeek, DeepSeek-V3.2 model cardchecked 2026-10-15 | - |
| OJBench (C++) | 30.4% | Moonshot AI, Kimi K2 Thinking release pagechecked 2026-10-15 | Effort: thinking; Tools: none |
Tool Use and Workflows
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Toolathlon | 38.6% | DeepSeek, DeepSeek-V3.2 model cardchecked 2026-10-15 | - |
| τ³-bench Banking | 22.4% | Mistral AI, Mistral Medium 3.5 model cardchecked 2026-10-15 | Effort: max; Harness: gpt-5.2 user simulator (low), 4 trials; terminal- or embedding-based agentic search retrieval, highest score |
| τ²-bench Telecom | 98% | Anthropic, Claude Sonnet 4.5checked 2026-10-15; printed on 4 more pages | Effort: extended thinking (up to 64K) |
| MCP Atlas | 43.8% | Anthropic, Claude Opus 4.5checked 2026-10-15; printed on 2 more pages | - |
| τ²-bench Retail | 86.2% | Anthropic, Claude Sonnet 4.5checked 2026-10-15; printed on 4 more pages | Effort: extended thinking (up to 64K) |
| τ²-bench Airline | 70% | Anthropic, Claude Sonnet 4.5checked 2026-10-15; printed on 1 more page | Effort: extended thinking (up to 64K) |
| τ²-bench (domain not stated) | 84.7% | DeepSeek, DeepSeek-V3.2 model cardchecked 2026-10-15; printed on 1 more page | - |
| τ³-bench Airline | 72% | Mistral AI, Mistral Medium 3.5 model cardchecked 2026-10-15 | Effort: max; Harness: gpt-5.2 user simulator (low), 4 trials; Sierra-reported for Claude Sonnet 4.5 and Qwen3.5 |
| τ³-bench Retail | 72.4% | Mistral AI, Mistral Medium 3.5 model cardchecked 2026-10-15 | Effort: max; Harness: gpt-5.2 user simulator (low), 4 trials; Sierra-reported for Claude Sonnet 4.5 and Qwen3.5 |
| τ³-bench Telecom | 84.9% | Mistral AI, Mistral Medium 3.5 model cardchecked 2026-10-15 | Effort: max; Harness: gpt-5.2 user simulator (low), 4 trials; Sierra-reported for Claude Sonnet 4.5 and Qwen3.5 |
Web Search and Research
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| BrowseComp | 43.9% | Anthropic, Claude Opus 4.6checked 2026-10-15; printed on 4 more pages | Effort: max; Tools: web search, web fetch, programmatic tool calling |
| BrowseComp-ZH | 42.4% | Moonshot AI, Kimi K2 Thinking release pagechecked 2026-10-15; printed on 1 more page | Effort: thinking; Tools: enabled; search, code interpreter, web browsing; Context: 256000 |
| Seal-0 | 53.4% | Moonshot AI, Kimi K2 Thinking release pagechecked 2026-10-15 | Effort: thinking; Tools: enabled; search, code interpreter, web browsing; Context: 256000 |
| FRAMES | 85% | Moonshot AI, Kimi K2 Thinking release pagechecked 2026-10-15; printed on 1 more page | Effort: thinking; Tools: enabled; search, code interpreter, web browsing; Context: 256000 |
| FinSearchComp T3 | 44% | Moonshot AI, Kimi K2 Thinking release pagechecked 2026-10-15 | Effort: thinking; Tools: enabled; search, code interpreter, web browsing; Context: 256000 |
| Research-Eval (Reka) | 41.2 | xAI, Grok 4.1 Fast and Agent Tools APIchecked 2026-10-15 | Tools: agentic search (web search, X search, code execution); exact tools per row not stated |
| X Browse | 14.6 | xAI, Grok 4.1 Fast and Agent Tools APIchecked 2026-10-15 | Tools: agentic search (web search, X search, code execution); exact tools per row not stated |
| FRAMES (average cost per task)lower is better | $0.078 | xAI, Grok 4.1 Fast and Agent Tools APIchecked 2026-10-15 | Tools: agentic search (web search, X search, code execution); exact tools per row not stated |
| Research-Eval (Reka, average cost per task)lower is better | $0.065 | xAI, Grok 4.1 Fast and Agent Tools APIchecked 2026-10-15 | Tools: agentic search (web search, X search, code execution); exact tools per row not stated |
| X Browse (average cost per task)lower is better | $0.126 | xAI, Grok 4.1 Fast and Agent Tools APIchecked 2026-10-15 | Tools: agentic search (web search, X search, code execution); exact tools per row not stated |
Instruction Following
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| IFBench | 57 | MiniMax, MiniMax-M2.5 model cardchecked 2026-10-15; printed on 1 more page | - |
| COLLIE | 90.5% | Mistral AI, Mistral Medium 3.5 model cardchecked 2026-10-15 | Effort: max |
Multilingual
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MMMLU | 89.1% | Anthropic, Claude Sonnet 4.5checked 2026-10-15; printed on 4 more pages | Effort: extended thinking (up to 128K) |
Writing and Conversation
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Longform Writing | 79.8 | Moonshot AI, Kimi K2 Thinking release pagechecked 2026-10-15 | Effort: thinking; Tools: none |
Healthcare
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| HealthBench | 44.2 | Moonshot AI, Kimi K2 Thinking release pagechecked 2026-10-15 | Effort: thinking; Tools: none |
| MedXpertQA MM | 54% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
| PMC-VQA | 55.9% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
| SLAKE | 73.6% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
Price
API list prices in US dollars per one million tokens, as printed on the developer's pricing page on the day checked.
| Price | Value | Printed by | Setup as printed |
|---|---|---|---|
| Input | $3 | Anthropic, Anthropic API pricingchecked 2026-10-15 | - |
| Cached input | $0.3 | Anthropic, Anthropic API pricingchecked 2026-10-15 | - |
| Cache write | $3.75 | Anthropic, Anthropic API pricingchecked 2026-10-15 | Cache: 5 minute cache |
| Cache write, 1 hour | $6 | Anthropic, Anthropic API pricingchecked 2026-10-15 | - |
| Output | $15 | Anthropic, Anthropic API pricingchecked 2026-10-15 | - |
| Input, batch | $1.5 | Anthropic, Anthropic API pricingchecked 2026-10-15 | - |
| Output, batch | $7.5 | Anthropic, Anthropic API pricingchecked 2026-10-15 | - |
More from Anthropic
- Claude Haiku 5.5
- Claude Sonnet 5.5
- Claude Opus 5.5
- Claude Fable 5.1
- Claude Mythos 5.1
- Claude Opus 5
- Claude Sonnet 5
- Claude Fable 5
- Claude Mythos 5
- Claude Opus 4.8
- Claude Opus 4.7
- Claude Mythos Preview
- Claude Sonnet 4.6
- Claude Opus 4.6
- Claude Opus 4.5
- Claude Haiku 4.5
- Claude Opus 4.1
- Claude Opus 4
- Claude Sonnet 4
- Claude 3.7 Sonnet
- Claude 3.5 Sonnet
Every number here is read from the same catalog file as the benchmark table, data version 2026103001. Where one page printed the model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.