whatmodel.sh / Claude Sonnet 5

Claude Sonnet 5

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Claude Sonnet 5 is a closed-weights language model from Anthropic, released June 2026.

74.9±0.5#43 of 719 LLMsfrom 42 benchmark results across 11 sources

$ how it compares

The LLMs ranked either side of Claude Sonnet 5, out of 828 with a score, then more from Anthropic.

#modelscorewhereRAMreleased
44Muse Spark Meta75±2comparison only2026-04
45Step 5 Preview StepFun75±1comparison only
46Claude Sonnet 5 Anthropic75±1api $4.00/M2026-06
47Grok 4.7 SpaceXAI (xAI)75±1api $3.00/M2026-09
48Claude Opus 4.6 Anthropic75±1api $10.00/M2026-02
more from Anthropic
2Claude Opus 5.5 Anthropic87±1api $8.00/M2026-09
4Claude Fable 5.1 Anthropic85±1api $20.00/M2026-09
5Claude Sonnet 5.5 Anthropic84±1api $4.00/M2026-09

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #36 of 300 · 38% of its score · 14 tests

Beats GPT-5.6 Luna, Gemini 3.7 Flash, Qwen3.8-Flash-Next here, though they rank higher overall. Trails GLM-5.3-Flash, Grok 4.7 here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
SWE-bench Pro (split unspecified) lab63.2% ✎▄█·▄·▄██···▄ beats 1/9very hard · avg 7.6%▇▇▇ 6.6%sharp · at its level
Terminal-Bench 4.0 +112.4% ✎█▅▅▅▅·▅▃▃▅▅▅ beats 1/20very hard · avg 0.5%▇▇ 5.5%sharp · several runs
Arena Agent (October 2026 aggregate IPS)0.0440 IPS ✎▁▁▁▄▂▃▇█▃▂▁▃ beats 38/49avg -0.312 IPS▇▇ 5.1%sharp
SWE-bench Multilingual lab78.3% ✎▃··▃·▅▃▃█··▃ beats 1/9hard · avg 31%▇ 3.4%typical
SWE-bench Verified lab85.2% ✎▁▁▁▁▁▁▂▄█▆▁▁ beats 127/131hard · avg 28%▇ 3.3%sharp
Terminal-Bench 2.1 lab80.4% ✎▁·▁·▁▁▄▃▄█▃▇ beats 39/56hard · avg 17%▇ 2.3%typical
GameLogicBench solve rate (72 Godot game-logic tasks, Claude Code scaffold)25.0% ✎█▄·█·▄·██▄·▄ beats 4/11very hard · avg 4.7%▇ 2.0%at its level
OSWorld-Verified lab81.2% ✎█▄·▂▆▂▄▂▆▄▆█ beats 20/25hard · avg 29%▇ 2.0%typical
6 more tests
testresultvs the others in its rundifficultycountswhy
E2E-SWE (build a working codebase from scratch, 186 tasks) pass@143.1% ✎▃··▅··▃█▃▅▃▅ beats 5/12very hard · avg 3.1%▇ 1.9%sharp · at its level
Humanity's Last Exam with tools lab +154.9% ✎▁······▁·▄·█ beats 2/10very hard · avg 14%▇ 1.6%at its level
Toolathlon (Tool Decathlon) lab54.3% ✎█···█··█·███ beats 1/5hard · avg 16%▇ 1.5%at its level
BrowseComp with context management lab84.7% ✎▅·▅▃·█▃·█▅·▃ beats 11/14medium · avg 44%▇ 1.4%typical
Toolathlon-Verified lab74.7% ✎▃·····█···▃▃ beats 0/5hard · avg 35%▇ 0.8%blunt · at its level
KlinikeBench strict pass@1 (333 interactive clinical encounters, Terminus-2)13.8% ✎▁▁·▁▁▁▁▅▃█▅█ beats 12/27very hard · avg 2.4%▇ 0.7%typical
coding▁▁▁▃▆██▇▆▄▃▁ #40 of 458 · 23% of its score · 7 tests

Beats GPT-5.5, GPT-5.6 Terra, GPT-5.4 here, though they rank higher overall. Trails Mistral Large 4 (public preview), Qwen3.8 27B, GLM-5.2 here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
SWE-bench Pro (split unspecified) lab63.2% ✎▄█·▄·▄██···▄ beats 1/9very hard · avg 7.6%▇▇▇ 6.6%sharp · at its level
Arena WebDev (Code Arena)1539.4 ✎▁▂▄▃▅█▄▇▆▄▁▁ beats 81/119avg 981▇▇ 4.5%sharp
SWE-bench Multilingual lab78.3% ✎▃··▃·▅▃▃█··▃ beats 1/9hard · avg 31%▇ 3.4%typical
SWE-bench Verified lab85.2% ✎▁▁▁▁▁▁▂▄█▆▁▁ beats 127/131hard · avg 28%▇ 3.3%sharp
GameLogicBench solve rate (72 Godot game-logic tasks, Claude Code scaffold)25.0% ✎█▄·█·▄·██▄·▄ beats 4/11very hard · avg 4.7%▇ 2.0%at its level
E2E-SWE (build a working codebase from scratch, 186 tasks) pass@143.1% ✎▃··▅··▃█▃▅▃▅ beats 5/12very hard · avg 3.1%▇ 1.9%sharp · at its level
Luau Leetcode fully solved (Torpedo Software)68.3% ✎▁▁▁▁·▁▂▃▅▃█▇ beats 21/49hard · avg 18%▇ 1.8%at its level
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #40 of 393 * · 2.5% of its score · 1 test

Beats GPT-6 Sol, MiMo-V2.6-Flash (309B-A15B), GPT-6 Luna here, though they rank higher overall. Trails MiMo-V2.5-Pro (1T-A42B), GLM-5.1, Grok 4.20 (0309 reasoning) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
LMArena (text)1461.8 ✎▁▁▁▂▃▃▃▆▆▇█▃ beats 332/379avg 1268▇ 2.5%blunt
math▁▂▃▅▆▅▄█▄▁▁▁ #34 of 398 · 23% of its score · 4 tests

Beats Claude Opus 4.7, Grok 4.5, Muse Spark 1.1 here, though they rank higher overall. Trails GPT-5.2, Qwen3.7 Max here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
FrontierMath tiers 1–3 v265.6% ✎▂··▁▁▃▄▄█▅▃▃ beats 52/82very hard · avg 1.2%▇▇▇▇▇ 12%sharp · at its level
FrontierMath tier 4 v229.3% ✎▁▂▁▂█▅▂▂▁▁▁▁ beats 34/63very hard · avg 0.1%▇▇▇▇ 10%sharp · at its level
USAMO 2026 lab79.5% ✎█·██······██ beats 2/4hard · avg 16%▇ 0.9%sharp · noisy
OTIS mock AIME 2024–202580.0% ✎▃▃▄▃▃▃▇▇█▃▂▅ beats 114/193very hard · avg 8.5%▇ 0.5%sharp · maxed out · noisy
vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #33 of 147 · 4.5% of its score · 4 tests

Beats MiMo-V2.6-Pro, MiMo-V2.6-Flash (309B-A15B), GPT-5.6 Luna here, though they rank higher overall. Trails Qwen3.6 Plus, Qwen3.7 Plus, Gemini 3 Flash (Preview) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Arena vision (style controlled)1264.5 ✎▁▁▂▂▂▂▄▅▅▆██ beats 113/145avg 1068▇ 1.7%blunt
CharXiv (reasoning questions) lab77.0% ✎███··█·██··█ beats 3/6hard · avg 29%▇ 1.6%typical
Arena document (style controlled)1475.7 ✎▁▁▃▃▄▃▇█▅▇▄▅ beats 26/37avg 1335▇ 0.8%blunt
SpaceCast-Bench predictive spatial reasoning (overall, 16 task types)38.3% ✎▃·▅▃▅▃··▅··█ beats 5/11very hard · avg 32%▇ 0.3%blunt · at its level · noisy
web search▂▁▂▆▄█▃▃▁▁·▁ #29 of 80 · 4.0% of its score · 3 tests

Beats Gemini 3.1 Pro (Preview), Claude Opus 4.7 here, though they rank higher overall. Trails GPT-4o mini (2024-07-18), Mistral Small 3.1 24B (2503), GPT-4o (2024-08-06) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Humanity's Last Exam with tools lab +154.9% ✎▁······▁·▄·█ beats 2/10very hard · avg 14%▇ 1.6%at its level
BrowseComp with context management lab84.7% ✎▅·▅▃·█▃·█▅·▃ beats 11/14medium · avg 44%▇ 1.4%typical
Arena search systems (style controlled)1196.1 ✎▁·▁▁▂▂▂█▅▂▁▁ beats 20/30avg 1055▇ 1.0%blunt
other tests15 not in a task · 24% of its score
testresultvs the others in its rundifficultycountswhy
KMMLU-HARD (Daekeun Kim, zero-shot)77.4% ✎█···▁▁▁▂▃·▆▂ beats 20/23very hard · avg 29%▇ 3.2%sharp · at its level
Mystery Game Puzzles (Epoch)35.0% ✎█··▂▃▇▄█▂▃▁▁ beats 56/74very hard · avg 9.4%▇ 3.2%sharp · noisy
Humanity's Last Exam lab +143.1% ✎▂·····▂·▄▂█▄ beats 2/10very hard · avg 5.8%▇ 3.1%at its level
Chess Puzzles (Epoch)16.0% ✎█▁▁▁▁▂▃▄▂▂▁▁ beats 70/135very hard · avg 5.9%▇ 2.1%sharp · at its level · noisy
GPQA Diamond80.3% ✎▁▁▁▁▂▄▆▄▅▇█▅ beats 129/208hard · avg 40%▇ 2.0%sharp
Global-MMLU lab +189.2% ✎▄▄····▄▄··▄█ beats 1/6medium · avg 63%▇ 1.6%typical
KoBALT-700 (Daekeun Kim)76.1% ✎▄▄█·····▄··▄ beats 1/5hard · avg 37%▇ 1.4%at its level
CLIcK Korean (Daekeun Kim)92.1% ✎▆▄▄▄█▄▂·▄█·▄ beats 20/23medium · avg 58%▇ 1.3%typical
7 more tests
testresultvs the others in its rundifficultycountswhy
KMMLU-Pro (Daekeun Kim)85.8% ✎▄·▄█·····▄·▄ beats 3/5medium · avg 40%▇ 1.3%typical
SimpleQA Verified33.7% ✎▁▁▃▂▃▃▆▆█▃▂▅ beats 26/78very hard · avg 8.0%▇ 1.0%at its level
HAE-RAE 1.0 (Daekeun Kim)92.0% ✎▄▄▄▂▂█▂▄▂█▂▄ beats 18/22medium · avg 63%▇ 1.0%maxed out
Icelandic LLM Leaderboard average (Miðeind, Sept 2026 suite)55.2% ✎▁▁▃▂▂▄▁▃▄▄▄█ beats 24/38very hard · avg 14%▇ 0.9%at its level
MuSR Korean (Daekeun Kim)81.6% ✎█·····███·██ beats 2/5medium · avg 42%▇ 0.8%typical
HealthBench lab59.2% ✎█·▄·····▄▄▄▄ beats 1/6hard · avg 29%▇ 0.7%blunt · at its level
RobloxQA v2.0 multiple choice (Torpedo Software)94.0% ✎▁▁▃▁█▄▃▃▆▂▂▂ beats 26/49easy · avg 78%▇ 0.6%maxed out

$ where to run it

Through an API

api · $4.00/M · $2 in / $10 out per 1M tokens · 1M context · docs.anthropic.com ↗ · price sourceStandard base token rate from models.dev. Provider-specific tiers, long-context surcharges, caching and promotions may differ.

$ how the score was worked out · Anthropic system cards and mo… 34%, Epoch AI — internal evaluatio… 31%, LMArena leaderboard dataset 16% and 8 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Anthropic system cards and model announcements (lab-reported scores) · lab-reported18▇▇▇▇▇▇▇▇▇▇ 34%Factual vendor-reported score (not licensed content)
Epoch AI — internal evaluations7▇▇▇▇▇▇▇▇▇ 31%CC BY 4.0
LMArena leaderboard dataset6▇▇▇▇▇ 16%CC BY 4.0
Daekeun Kim Korean LLM benchmark runs6▇▇▇ 9%MIT
Terminal-Bench 4.0 official leaderboard submissions1▇ 2%Apache-2.0
Roblox LLM Leaderboard (Torpedo Software): RobloxQA v2.0 and Luau Leetcode2▇ 2%MIT
GameLogicBench: Evaluating Coding Agents on Runtime Game Logic with Tick-Level State Assertions (arXiv:2609.21562v2)1▇ 2%CC BY 4.0
E2E-SWE: Benchmarking LLMs on Building Working Codebases from Scratch (arXiv:2609.38335v1)1▇ 2%CC BY 4.0
Miðeind Icelandic LLM Leaderboard1▇ 1%CC BY 4.0
KlinikeBench: Evaluating Language Models Beyond Diagnostic Accuracy (arXiv:2609.38480v2)1▇ 1%CC BY 4.0
SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models (arXiv:2610.12402v1)1▇ 0%CC BY 4.0
prior · pull toward the middle▇ 0%
weighted average= 74.9

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json