AI leaderboard · Agent work
Best AI for agents: 57 models ranked on measured results
LMArena’s Agent Arena: people give an agent real work on arena.ai, and the arena measures how the work goes with each model in charge: tasks the user marks as done, praise against complaints, corrections carried out, shell commands and calls to tools that do not exist.
Run Claude Opus 5.5 up to 60% below list price. $25 a month brings $110 of API credit and cancels any time; $25 once buys $100. Scores from LMArena’s Agent Arena, published 8 November 2026, under CC BY 4.0. Prices from each maker’s pricing page, read on 12 November 2026.
+20.0%Claude Opus 5.5 from Anthropic
- Models ranked
- 57
- Level with the leader
- 1
- On sale with AIGROW credit
- 8
- Data of
- 8 November 2026
Level: its interval overlaps the leader’s.
On LMArena’s Agent Arena, published 8 November 2026, Claude Opus 5.5 from Anthropic leads 57 models with a score of +20.0% from real agent sessions. One other model is statistically level with it. AIGROW API credit sells it at $1.60 / $8.00 instead of $4.00 / $20.00 per million input / output tokens.
Run models from this ranking below list price
AIGROW API credit runs these models through one key, billed per token at the rates below. $25 a month brings $110 of API credit and cancels any time; $25 once buys $100.
- No. 1Claude Opus 5.5Anthropic
List price $4.00 / $20.00$1.60 / $8.00with AIGROW API creditRun Claude Opus 5.5 up to 60% below list priceCertified genuine by AIGROW on 11 November 2026 - No. 3Claude Sonnet 5.5Anthropic
List price $2.00 / $10.00$0.80 / $4.00with AIGROW API creditRun Claude Sonnet 5.5 up to 60% below list priceCertified genuine by AIGROW on 11 November 2026 - No. 4Claude Fable 5.1Anthropic
List price $10.00 / $50.00$4.00 / $20.00with AIGROW API creditRun Claude Fable 5.1 up to 60% below list priceCertified genuine by AIGROW on 11 November 2026
US dollars per million input / output tokens. Certified genuine by AIGROW means the model passed AIGROW’s own check against the same model at its maker on 11 November 2026. No model maker reviewed or endorsed it.
Best AI for agents
The score is how much better tasks went than with the average model in the arena, not a share of tasks finished, and it comes from the arena’s own users and tools.
| # | Model | Score | Sessions | Median cost per task | Per 1M tokens, in / out |
|---|---|---|---|---|---|
| 1 | Claude Opus 5.5Anthropic · Claude Opus 5.5 (Max)Run Claude Opus 5.5 up to 60% below list priceCertified genuine by AIGROW on 11 November 2026 | 95% interval +17.9% to +22.0% | 14,358 | $6.31 | |
| 2 | Gemini 4 CarbonGoogle · Gemini 4 Carbon (High) | 95% interval +15.9% to +17.3% | 53,721 | $1.54 | $2.00 / $10.00 |
| 3 | Claude Sonnet 5.5Anthropic · Claude Sonnet 5.5 (Max)Run Claude Sonnet 5.5 up to 60% below list priceCertified genuine by AIGROW on 11 November 2026 | 95% interval +13.2% to +18.8%Its interval overlaps 2 models above it | 8,670 | $3.42 | |
| 4 | Claude Fable 5.1Anthropic · Claude Fable 5.1 (Max)Run Claude Fable 5.1 up to 60% below list priceCertified genuine by AIGROW on 11 November 2026 | 95% interval +12.4% to +16.3%Its interval overlaps 2 models above it | 17,468 | $6.87 | |
| 5 | GPT-6 AstraOpenAI · GPT-6 Astra (Max)Run GPT-6 Astra up to 75% below list priceCertified genuine by AIGROW on 11 November 2026 | 95% interval +12.2% to +16.3%Its interval overlaps 3 models above it | 15,520 | $7.04 | |
| 6 | GPT-6.1 SolOpenAI · GPT-6.1 Sol (Max)Run GPT-6.1 Sol up to 75% below list priceCertified genuine by AIGROW on 11 November 2026 | 95% interval +11.4% to +13.6%Its interval overlaps 3 models above it | 30,514 | $0.98 | |
| 7 | GPT-6.1 TerraOpenAI · GPT-6.1 Terra (Max)Run GPT-6.1 Terra up to 75% below list priceCertified genuine by AIGROW on 11 November 2026 | 95% interval +10.8% to +13.5%Its interval overlaps 4 models above it | 23,276 | $0.80 | |
| 8 | Gemini 4 ArgonGoogle · Gemini 4 Argon (High)Run Gemini 4 Argon up to 50% below list priceCertified genuine by AIGROW on 11 November 2026 | 95% interval +10.0% to +11.7%Its interval overlaps 2 models above it | 53,301 | $0.82 | |
| 9 | Muse Wonder Full (Max)Meta | 95% interval +8.2% to +10.0%Its interval overlaps 1 model above it | 32,477 | Not published | No list price |
| 10 | Claude Fable 5 (High)Anthropic | 95% interval +6.6% to +8.8%Its interval overlaps 1 model above it | 51,053 | $3.42 | No list price |
| 11 | Claude Opus 5Anthropic · Claude Opus 5 (Max) | 95% interval +6.2% to +8.8%Its interval overlaps 2 models above it | 31,928 | $4.14 | $5.00 / $25.00 |
| 12 | GPT-6 SolOpenAI · GPT-6 Sol (Max) | 95% interval +5.5% to +8.6%Its interval overlaps 3 models above it | 12,836 | $1.76 | $2.00 / $10.00 |
| 13 | Kimi K3.1 ProMoonshot AI · Kimi K3.1 Pro (Max) · open weights | 95% interval +4.1% to +6.4%Its interval overlaps 2 models above it | 22,581 | $1.31 | $2.00 / $10.00 |
| 14 | Hy4Tencent | 95% interval +3.2% to +6.1%Its interval overlaps 2 models above it | 17,888 | $0.59 | No list price |
| 15 | GPT-5.6 SolOpenAI · GPT-5.6 Sol (xHigh) | 95% interval +3.5% to +5.6%Its interval overlaps 3 models above it | 45,371 | $1.72 | $4.00 / $20.00 |
| 16 | Grok 4.7xAI · Grok 4.7 (xHigh)Run Grok 4.7 up to 75% below list priceCertified genuine by AIGROW on 11 November 2026 | 95% interval +1.2% to +4.1%Its interval overlaps 3 models above it | 18,848 | $2.91 | |
| 17 | Gemini 3.8 FlashGoogle · Gemini 3.8 Flash (High) | 95% interval +1.5% to +3.2%Its interval overlaps 2 models above it | 52,722 | $0.75 | $0.75 / $3.75 |
| 18 | MiMo-V2.6-ProXiaomi · open weights | 95% interval +1.4% to +3.2%Its interval overlaps 3 models above it | 49,297 | $0.17 | No list price |
| 19 | Muse Spark 1.3Meta · Muse Spark 1.3 (Max) | 95% interval +1.4% to +3.0%Its interval overlaps 3 models above it | 80,655 | $0.47 | $1.25 / $4.25 |
| 20 | DeepSeek V4.1 FlashDeepSeek · DeepSeek-V4.1-Flash-20261017 (Max) · open weights | 95% interval +1.5% to +2.6%Its interval overlaps 4 models above it | 78,504 | $0.15 | $0.30 / $1.20 |
| 21 | Claude Sonnet 5Anthropic · Claude Sonnet 5 (High) | 95% interval +0.9% to +3.2%Its interval overlaps 6 models above it | 41,726 | $0.91 | $2.00 / $10.00 |
| 22 | Muse WonderMeta · Muse Wonder (Max) | 95% interval +0.6% to +3.3%Its interval overlaps 7 models above it | 13,921 | $1.28 | $4.00 / $20.00 |
| 23 | GPT-6 LunaOpenAI · GPT-6 Luna (Max) | 95% interval +1.1% to +2.0%Its interval overlaps 7 models above it | 228,261 | $0.09 | $0.10 / $0.50 |
| 24 | GLM-5.3Z.ai · GLM-5.3 (Max) · open weights | 95% interval +0.2% to +1.8%Its interval overlaps 8 models above it | 95,597 | $1.09 | $1.40 / $4.40 |
| 25 | Kimi K3Moonshot AI · Kimi K3 (Max) · open weights | 95% interval 0.0% to +1.2%Its interval overlaps 4 models above it | 160,495 | $0.93 | $3.00 / $15.00 |
| 26 | Hy4 previewTencent · open weights | 95% interval -0.3% to +1.2%Its interval overlaps 5 models above it | 102,097 | $0.46 | No list price |
| 27 | Claude Haiku 5.5Anthropic · Claude Haiku 5.5 (High) | 95% interval -0.6% to +1.5%Its interval overlaps 9 models above it | 31,731 | $0.08 | $0.10 / $0.50 |
| 28 | GLM-5.2 (Max)Z.ai · open weights | 95% interval -0.4% to +1.2%Its interval overlaps 8 models above it | 98,965 | $0.90 | No list price |
| 29 | Qwen3.8 MaxAlibaba · Qwen3.8-Max | 95% interval -0.3% to +1.0%Its interval overlaps 7 models above it | 80,457 | $1.24 | $2.00 / $6.00 |
| 30 | MiMo-V2.6-FlashXiaomi · open weights | 95% interval -1.0% to +1.2%Its interval overlaps 10 models above it | 32,025 | $0.07 | No list price |
| 31 | Inkling 2 SmallThinky | 95% interval -2.5% to +2.6%Its interval overlaps 15 models above it | 4,063 | $0.26 | No list price |
| 32 | Grok 4.6 (xHigh)xAI | 95% interval -1.4% to +1.1%Its interval overlaps 10 models above it | 30,247 | $1.46 | No list price |
| 33 | GPT-5.6 TerraOpenAI · GPT-5.6 Terra (xHigh) | 95% interval -2.7% to -1.1%Its interval overlaps 2 models above it | 44,425 | $0.64 | $2.00 / $12.00 |
| 34 | Gemini 3.7 Flash (High)Google | 95% interval -2.8% to -1.0%Its interval overlaps 4 models above it | 63,336 | $0.52 | No list price |
| 35 | GPT-6 LunaOpenAI · GPT-6 Luna (Medium) | 95% interval -2.5% to -1.5%Its interval overlaps 3 models above it | 76,341 | $0.03 | $0.10 / $0.50 |
| 36 | MiniMax-M3.1-FlashMiniMax · open weights | 95% interval -4.3% to -1.3%Its interval overlaps 5 models above it | 13,089 | $0.24 | No list price |
| 37 | GLM-5.3-FlashZ.ai · open weights | 95% interval -3.5% to -2.1%Its interval overlaps 5 models above it | 95,526 | $0.17 | $0.15 / $0.50 |
| 38 | Step 5 PreviewStepFun | 95% interval -4.8% to -2.0%Its interval overlaps 6 models above it | 17,051 | $0.66 | No list price |
| 39 | Gemini 3.1 Pro (preview)Google · Gemini 3.1 Pro Preview | 95% interval -5.8% to -2.9%Its interval overlaps 3 models above it | 87,553 | $0.55 | $2.00 / $12.00 |
| 40 | Qwen3.8 27BAlibaba · Qwen3.8-27B · open weights | 95% interval -5.5% to -3.7%Its interval overlaps 3 models above it | 63,984 | $0.40 | $0.50 / $3.00 |
| 41 | Qwen3.8-Flash-NextAlibaba · open weights | 95% interval -5.4% to -3.9%Its interval overlaps 4 models above it | 109,992 | $0.23 | No list price |
| 42 | Mistral Large 4Mistral AI · open weights | 95% interval -8.0% to -4.2%Its interval overlaps 5 models above it | 5,559 | $0.32 | $0.70 / $2.10 |
| 43 | GPT-6 LunaOpenAI · GPT-6 Luna (Low) | 95% interval -6.9% to -5.5%Its interval overlaps 2 models above it | 43,861 | $0.01 | $0.10 / $0.50 |
| 44 | Beam 1.0 501B A23BReflection · open weights | 95% interval -9.0% to -3.9%Its interval overlaps 7 models above it | 12,257 | Not published | No list price |
| 45 | Qwen3.7-MaxAlibaba | 95% interval -8.4% to -5.9%Its interval overlaps 3 models above it | 43,146 | $0.35 | No list price |
| 46 | Hy3Tencent · open weights | 95% interval -9.1% to -6.2%Its interval overlaps 4 models above it | 32,686 | $0.08 | No list price |
| 47 | Muse Spark 1.1Meta | 95% interval -8.5% to -7.2%Its interval overlaps 4 models above it | 149,186 | Not published | No list price |
| 48 | Qwen3.7 PlusAlibaba · Qwen3.7-Plus | 95% interval -9.4% to -6.3%Its interval overlaps 6 models above it | 24,539 | $0.13 | $0.40 / $1.60 |
| 49 | Muse Spark 1.2 (xHigh)Meta | 95% interval -8.9% to -7.2%Its interval overlaps 6 models above it | 54,770 | $0.41 | No list price |
| 50 | Spark-X2.6Synlan | 95% interval -11.6% to -9.1%Its interval overlaps 2 models above it | 11,434 | $0.09 | No list price |
| 51 | MiniMax M3MiniMax · MiniMax-M3 · open weights | 95% interval -11.6% to -9.2%Its interval overlaps 2 models above it | 44,355 | $0.32 | $0.30 / $1.20 |
| 52 | Mistral Medium 3.5Mistral AI · open weights | 95% interval -15.7% to -12.1% | 13,935 | $0.40 | $1.50 / $7.50 |
| 53 | Nemotron 3 UltraNvidia · open weights | 95% interval -17.7% to -11.1%Its interval overlaps 3 models above it | 13,691 | $0.12 | No list price |
| 54 | NVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4)Nvidia · open weights | 95% interval -17.0% to -15.5%Its interval overlaps 2 models above it | 29,893 | Not published | No list price |
| 55 | InklingThinky · open weights | 95% interval -18.6% to -15.5%Its interval overlaps 3 models above it | 44,032 | $0.34 | No list price |
| 56 | Inkling-SmallThinky · open weights | 95% interval -19.7% to -15.7%Its interval overlaps 4 models above it | 14,747 | $0.18 | No list price |
| 57 | Solar Pro 4Upstage | 95% interval -25.0% to -20.6% | 11,012 | $0.04 | No list price |
How to read this ranking
What the scores measure, what they leave out, and where each number comes from.
- What it measures
- LMArena’s Agent Arena: people give an agent real work on arena.ai, and the arena measures how the work goes with each model in charge: tasks the user marks as done, praise against complaints, corrections carried out, shell commands and calls to tools that do not exist.
- What it leaves out
- The score is how much better tasks went than with the average model in the arena, not a share of tasks finished, and it comes from the arena’s own users and tools.
- How to read the margin
- Each score carries LMArena’s 95% interval. Where two models’ intervals overlap, the sessions cannot say which is better. Rows are still numbered by score, and a row says how many models above it overlap.
- Cost per task
- The median cost of a task in the arena’s sessions, in US dollars, as LMArena records it. It depends on the model’s price and on how many tokens it spends.
- Not updated in over 120 days
- A result run more than 120 days before the newest data on these pages carries this mark, so an older result can be weighed as one.
- Prices
- The list price per million tokens from the maker’s own pricing page, read on 12 November 2026. It is a price per token, not per answer: a higher reasoning setting spends more tokens on the same question. Where AIGROW API credit sells a model, the list price is struck through and the price through the credit sits under it.
Source: LMArena, leaderboard dataset (revision 4217986c, published 8 November 2026), licensed under CC BY 4.0. We change no score: we add the prices and the links to our model pages.
What each of them costs to run
The price list works three everyday jobs on every model, and our own benchmark runs them on real business tasks to see which pass.
Data of 8 November 2026