LLM Benchmark

LLM Benchmark ranks LLMs on public LLM benchmarks for coding, agents, reasoning and math, with API prices and cost per solved task. Updated daily.

View on llmbenchmark.ai

Hosted MCP Server

npx add-mcp 'https://api.llmbenchmark.ai/mcp'

Installs into Claude Code, Codex, Cursor and more

Documentation

Current leaders

The top three overall today, each with its place on every category board, among every model measured there. Full leaderboard

  1. 1

    Claude Opus 5.5

    Anthropic 86.9 LLMBenchmark Index ±2.3, leads by 3.0
    • 1 Coding 88.9 #1/100
    • 2 Agentic 82.0 #2/47
    • 2 Reasoning 90.6 #2/119
    • 3 Math 93.7 #3/106
    • 4 Knowledge 85.0 #4/70
    • 5 Web design 1326 #5/134 $20.00 per M output, 1M context
  2. 2

    GPT-6 Astra

    OpenAI 83.9 LLMBenchmark Index ±3.6, 3.0 behind the leader
    • 1 Knowledge 87.3 #1/70
    • 3 Coding 85.1 #3/100
    • 3 Agentic 81.0 #3/47
    • 3 Web design 1338 #3/134
    • 4 Reasoning 89.7 #4/119
    • 5 Math 93.1 #5/106 $50.00 per M output, 1.05M context
  3. 3

    Claude Fable 5.1

    Anthropic 83.5 LLMBenchmark Index ±3.9, 3.4 behind the leader
    • 1 Agentic 82.2 #1/47
    • 2 Math 93.8 #2/106
    • 4 Coding 84.3 #4/100
    • 6 Reasoning 88.1 #6/119
    • 6 Knowledge 82.2 #6/70
    • 11 Web design 1312 #11/134 $50.00 per M output, 1M context

Find your model

Three questions, five answers. Every pick is ranked by a public board, and a model with no number for what you chose is left out and counted, not guessed.

LLMBenchmark Index

  1. 1 Claude Opus 5.5 Anthropic 86.9 $20.00 / M out · 1M context · 96 tok/s
  2. 2 GPT-6 Astra OpenAI 83.9 $50.00 / M out · 1.05M context · 65 tok/s
  3. 3 Claude Fable 5.1 Anthropic 83.5 $50.00 / M out · 1M context · 62 tok/s
  4. 4 Claude Opus 5 Anthropic 82.7 $25.00 / M out · 1M context
  5. 5 Gemini 4 Argon Google 82.6 $10.00 / M out · 1.05M context · 80 tok/s

129 models fit; 18 of them have no number for this and are left out.

Quality against price

Every paid model on the overall index. Up is better, left is cheaper; the dashed line is the best you can get at each price.

0 10 20 30 40 50 60 70 80 90 $0.1 $0.3 $1 $3 $10 $30 median $3.12 median 57.9 Better and cheaper than the median: 15 Better, dearer: 40 Weaker, cheaper: 40 Weaker and dearer than the median: 14 Claude Opus 5.5 (Anthropic): overall 86.9, $20.00 per million output tokens GPT-6 Astra (OpenAI): overall 83.9, $50.00 per million output tokens Claude Fable 5.1 (Anthropic): overall 83.5, $50.00 per million output tokens Claude Opus 5 (Anthropic): overall 82.7, $25.00 per million output tokens Gemini 4 Argon (Google): overall 82.6, $10.00 per million output tokens GPT-6.1 Sol (OpenAI): overall 82.0, $10.00 per million output tokens Claude Sonnet 5.5 (Anthropic): overall 81.7, $10.00 per million output tokens Claude Fable 5 (Anthropic): overall 78.9, $50.00 per million output tokens GPT-6.1 Terra (OpenAI): overall 76.8, $5.00 per million output tokens GPT-6 Sol (OpenAI): overall 76.4, $10.00 per million output tokens GPT-5.5 (OpenAI): overall 75.7, $30.00 per million output tokens GPT-5.6 Sol (OpenAI): overall 75.5, $20.00 per million output tokens Kimi K3.1 Pro (Moonshot AI): overall 74.4, $10.00 per million output tokens Claude Opus 4.7 (Anthropic): overall 74.4, $25.00 per million output tokens Muse Spark 1.1 (Meta): overall 73.3, $4.25 per million output tokens GPT-5.4 (OpenAI): overall 72.6, $15.00 per million output tokens Gemini 3.6 Flash (Google): overall 72.5, $3.75 per million output tokens Claude Opus 4.6 (Anthropic): overall 70.4, $25.00 per million output tokens DeepSeek V4.1 Flash (DeepSeek): overall 70.2, $0.9085 per million output tokens DeepSeek V4.1 Flash 1017 (DeepSeek): overall 68.6, $0.60 per million output tokens Kimi K3 (Moonshot AI): overall 68.4, $15.00 per million output tokens GPT-5.6 Terra (OpenAI): overall 68.2, $12.00 per million output tokens Claude Opus 4.8 (Anthropic): overall 68.2, $25.00 per million output tokens Qwen3.8 Max (0902) (Qwen): overall 67.6, $6.00 per million output tokens Qwen3.7 Max (Qwen): overall 67.1, $4.425 per million output tokens Qwen3.5 397B A17B (Qwen): overall 67.1, $2.34 per million output tokens Grok 4.6 (xAI): overall 67.0, $6.00 per million output tokens GLM 5.3 (Z.ai): overall 66.9, $4.40 per million output tokens Gemini 3.7 Flash (Google): overall 66.4, $3.75 per million output tokens Gemma 4 26B A4B (Google): overall 66.3, $0.60 per million output tokens GLM 5.3 Flash (Z.ai): overall 65.8, $0.50 per million output tokens GLM 5.2 (Z.ai): overall 65.6, $4.40 per million output tokens Claude Sonnet 4.6 (Anthropic): overall 65.6, $15.00 per million output tokens Gemini 3.8 Flash (Google): overall 65.5, $3.75 per million output tokens Kimi K2.6 (Moonshot AI): overall 65.2, $4.00 per million output tokens GPT-5.2 (OpenAI): overall 64.4, $14.00 per million output tokens Claude Haiku 5.5 (Anthropic): overall 64.2, $0.50 per million output tokens Grok 4.7 (xAI): overall 63.5, $6.00 per million output tokens Muse Spark 1.3 (Meta): overall 63.4, $4.25 per million output tokens Kimi K2.7 Code (Moonshot AI): overall 63.3, $4.00 per million output tokens Qwen3.5-35B-A3B (Qwen): overall 63.2, $1.30 per million output tokens MiMo-V2.6-Pro (Xiaomi): overall 62.7, $0.87 per million output tokens gpt-oss-120b (OpenAI): overall 62.5, $0.50 per million output tokens GLM 4.7 (Z.ai): overall 62.3, $2.20 per million output tokens GLM 5.1 (Z.ai): overall 61.9, $4.40 per million output tokens GPT-5.6 Luna (OpenAI): overall 61.1, $1.20 per million output tokens Claude Sonnet 5 (Anthropic): overall 60.8, $10.00 per million output tokens Grok 4.5 (xAI): overall 60.3, $6.00 per million output tokens Qwen3.6 Plus (Qwen): overall 60.0, $1.95 per million output tokens Gemini 3 Flash Preview (Google): overall 59.7, $3.00 per million output tokens GPT-5.1 (OpenAI): overall 59.2, $10.00 per million output tokens Gemini 3.5 Flash (Google): overall 59.0, $9.00 per million output tokens Claude Opus 4.5 (Anthropic): overall 58.7, $25.00 per million output tokens Qwen3.7 Plus (Qwen): overall 58.5, $1.28 per million output tokens DeepSeek V4 Pro 0813 (DeepSeek): overall 58.0, $1.98 per million output tokens GLM 5 (Z.ai): overall 57.9, $3.20 per million output tokens o3 (OpenAI): overall 57.8, $8.00 per million output tokens R1 0528 (DeepSeek): overall 57.4, $2.18 per million output tokens Grok 4.3 (xAI): overall 56.8, $2.50 per million output tokens Muse Spark 1.2 (Meta): overall 56.7, $4.25 per million output tokens DeepSeek V4 Flash 0731 (DeepSeek): overall 56.6, $0.348 per million output tokens GPT-6 Luna (OpenAI): overall 56.5, $0.50 per million output tokens Step 5 Preview (StepFun): overall 56.0, $2.70 per million output tokens Gemini 3.1 Pro Preview (Google): overall 56.0, $12.00 per million output tokens MiMo-V2.6-Flash (Xiaomi): overall 55.9, $0.28 per million output tokens GPT-5 (OpenAI): overall 55.5, $10.00 per million output tokens Qwen3.8 27B (Qwen): overall 55.4, $2.55 per million output tokens MiniMax M3.1 Flash (MiniMax): overall 54.9, $0.60 per million output tokens Claude Sonnet 4.5 (Anthropic): overall 53.8, $15.00 per million output tokens Qwen3 30B A3B Instruct 2507 (Qwen): overall 53.0, $0.30 per million output tokens Qwen3 32B (Qwen): overall 52.2, $0.425 per million output tokens Mistral Large 4 (Mistral AI): overall 51.4, $4.18 per million output tokens DeepSeek V4 Pro 0423 (DeepSeek): overall 51.4, $3.12 per million output tokens Qwen3.6 27B (Qwen): overall 50.2, $2.70 per million output tokens Qwen3.5-Flash (Qwen): overall 50.1, $0.26 per million output tokens GLM 4.7 Flash (Z.ai): overall 50.0, $0.40 per million output tokens Nemotron 3 Ultra (NVIDIA): overall 49.9, $2.40 per million output tokens GPT-5.4 Nano (OpenAI): overall 49.2, $1.25 per million output tokens Qwen3 30B A3B (Qwen): overall 48.4, $0.50 per million output tokens Kimi K2.5 (Moonshot AI): overall 47.6, $2.85 per million output tokens gpt-oss-20b (OpenAI): overall 47.6, $0.145 per million output tokens MiniMax M3 (MiniMax): overall 47.2, $1.20 per million output tokens Gemini 3.5 Flash Lite (Google): overall 46.8, $2.50 per million output tokens GPT-5 Mini (OpenAI): overall 46.7, $2.00 per million output tokens Gemini 2.5 Pro (Google): overall 46.5, $10.00 per million output tokens Claude Haiku 4.5 (Anthropic): overall 46.3, $5.00 per million output tokens GPT-5.4 Mini (OpenAI): overall 45.7, $4.50 per million output tokens Gemma 4 31B (Google): overall 45.4, $0.40 per million output tokens Inkling (Thinking Machines): overall 43.5, $4.05 per million output tokens Muse Spark Open (Meta): overall 42.5, $4.25 per million output tokens GPT-4.1 (OpenAI): overall 41.9, $8.00 per million output tokens Inkling Small (Thinking Machines): overall 40.9, $1.20 per million output tokens GPT-5 Nano (OpenAI): overall 40.8, $0.40 per million output tokens Gemma 3 27B (Google): overall 40.6, $0.30 per million output tokens DeepSeek V3.2 (DeepSeek): overall 40.0, $0.80 per million output tokens MiMo-V2.5-Pro (Xiaomi): overall 39.9, $0.87 per million output tokens MiniMax M2.5 (MiniMax): overall 38.7, $1.20 per million output tokens GPT-4.1 Mini (OpenAI): overall 37.0, $1.60 per million output tokens Gemma 3 12B (Google): overall 36.4, $0.15 per million output tokens GPT-4o (2024-05-13) (OpenAI): overall 33.4, $15.00 per million output tokens Qwen2.5 72B Instruct (Qwen): overall 32.2, $0.40 per million output tokens Llama 3.3 70B Instruct (Meta Llama): overall 31.6, $0.715 per million output tokens GPT-4o (2024-08-06) (OpenAI): overall 30.9, $10.00 per million output tokens Llama 4 Scout (Meta Llama): overall 30.8, $0.50 per million output tokens Phi 4 (Microsoft): overall 30.8, $0.14 per million output tokens Nemotron 3.5 Super VL (NVIDIA): overall 30.4, $0.50 per million output tokens Llama 3.1 70B Instruct (Meta Llama): overall 29.9, $0.56 per million output tokens GPT-4.1 Nano (OpenAI): overall 28.8, $0.42 per million output tokens Gemma 3 4B (Google): overall 28.5, $0.075 per million output tokens Llama 3.1 8B Instruct (Meta Llama): overall 20.9, $0.08 per million output tokens Llama 3.2 1B Instruct (Meta Llama): overall 10.7, $0.201 per million output tokens Claude Opus 5.5 Gemini 4 Argon GPT-6.1 Terra Muse Spark 1.1 Gemini 3.6 Flash DeepSeek V4.1 Flash Output price per million tokens, log scale Overall index

  • OpenAI
  • Qwen
  • Google
  • Mistral AI
  • DeepSeek
  • Anthropic
  • MiniMax
  • Meta, Meta Llama
  • NVIDIA
  • Xiaomi
  • StepFun
  • Microsoft
  • Moonshot AI, xAI, Z.ai
  • Other makers

111 paid models. A dot below the line has a model at least as good for the same money or less. Hover a dot for its numbers; click it for the model page. Overall: LLMBenchmark Index. Price: OpenRouter.

The top two, category by category

The two highest models overall. Every pair of ranked models has a page like this; build one or open a popular one.

  • 86.9 LLMBenchmark Index 83.9
  • 58.4 AA Intelligence Index 56.9
  • 88.9 Coding 85.1
  • 82.0 Agentic 81.0
  • 90.6 Reasoning 89.7
  • 93.7 Math 93.1
  • 85.0 Knowledge 87.3
  • 1326 Web design 1338

The big number is categories won, counted only where both have a result.Full comparison

What a working answer costs

Our own metric, published nowhere else: one SWE-bench Verified attempt divided by the share that succeed. Known for 84 models. How it is computed

Why a cheap token is not a cheap answer

Cost per solved task is our own number: what one SWE-bench Verified attempt costs, divided by the share of attempts that succeed. GPT-4.1 Mini costs 5.0x less per token than GPT-4.1 and 4.9x more per solved task.

Price per million output tokensGPT-4.1 Mini 5.0x cheaper

GPT-4.1 $8.00

GPT-4.1 Mini $1.60

Cost per solved taskGPT-4.1 Mini 4.9x dearer

GPT-4.1 $0.37

GPT-4.1 Mini $1.818

The five cheapest working answers

  1. 1 DeepSeek V4 Flash 0731 resolves 88.8% $0.01
  2. 2 MiMo-V2.5 resolves 71.0% $0.01
  3. 3 GLM 5.3 Flash resolves 92.0% $0.02
  4. 4 MiMo-V2.5-Pro resolves 74.0% $0.03
  5. 5 GPT-5.6 Luna resolves 93.0% $0.04

See every model on the cost and quality chart

Top five in every category

One row per public board, shown as published: the five best at their own results, the leader with its lead over second. All leaderboards · Best for a job

  1. Coding Vibe Coding Index · 100 measured
    1. 1 Claude Opus 5.5 Leads by 1.6 points 88.9 2. 2 Claude Sonnet 5.5 87.3 3. 3 GPT-6 Astra 85.1 4. 4 Claude Fable 5.1 84.3 5. 5 GPT-6.1 Sol 82.3
  2. Agentic Agentic Index · 47 measured
    1. 1 Claude Fable 5.1 Leads by 0.2 points 82.2 2. 2 Claude Opus 5.5 82.0 3. 3 GPT-6 Astra 81.0 4. 4 Claude Sonnet 5.5 78.2 5. 5 Claude Opus 5 77.6
  3. Reasoning Reasoning Index · 119 measured
    1. 1 Gemini 4 Argon Leads by 0.1 points 90.7 2. 2 Claude Opus 5.5 90.6 3. 3 GPT-5.4 Pro 90.1 4. 4 GPT-6 Astra 89.7 5. 5 GPT-6.1 Sol 88.6
  4. Math Math Index · 106 measured
    1. 1 GPT-6.1 Terra Leads by 1.6 points 95.4 2. 2 Claude Fable 5.1 93.8 3. 3 GPT-6.1 Sol 93.7 4. 4 Claude Opus 5.5 93.7 5. 5 GPT-6 Astra 93.1
  5. Knowledge Knowledge Index · 70 measured
    1. 1 GPT-6 Astra Leads by 0.6 points 87.3 2. 2 Gemini 4 Argon 86.7 3. 3 GPT-6.1 Sol 85.5 4. 4 Claude Opus 5.5 85.0 5. 5 Gemini 3.1 Pro Preview 82.6
  6. Web design Design Arena: websites · 134 measured
    1. 1 Muse Wonder Leads by 11 Elo 1366 2. 2 Muse Spark 1.3 1355 3. 3 GPT-6 Astra 1338 4. 4 Kimi K3 1335 5. 5 Claude Opus 5.5 1326

The makers' medal table

In every category the makers are placed by their best model: gold, silver, bronze. Equal scores share a place, and a maker with no model on a board says so.

Maker123LLMBenchmark IndexCodingAgenticReasoningMathKnowledgeWeb design
Anthropic3211 86.9 Claude Opus 5.51 88.9 Claude Opus 5.51 82.2 Claude Fable 5.12 90.6 Claude Opus 5.52 93.8 Claude Fable 5.13 85.0 Claude Opus 5.54 1326 Claude Opus 5.5
OpenAI2412 83.9 GPT-6 Astra2 85.1 GPT-6 Astra2 81.0 GPT-6 Astra3 90.1 GPT-5.4 Pro1 95.4 GPT-6.1 Terra1 87.3 GPT-6 Astra2 1338 GPT-6 Astra
Google1133 82.6 Gemini 4 Argon4 81.0 Gemini 4 Argon3 72.4 Gemini 4 Argon1 90.7 Gemini 4 Argon3 92.3 Gemini 4 Argon2 86.7 Gemini 4 Argon5 1316 Gemini 4 Argon
Meta1005 73.3 Muse Spark 1.16 72.0 Muse Wonder11 35.2 Muse Spark 1.38 77.6 Muse Spark 1.35 86.1 Muse Spark 1.35 70.8 Muse Spark 1.21 1366 Muse Wonder
Moonshot AI0024 74.4 Kimi K3.1 Pro3 82.2 Kimi K3.1 Pro4 65.5 Kimi K3.1 Pro4 81.4 Kimi K3.1 Pro7 84.2 Kimi K37 68.5 Kimi K33 1335 Kimi K3
DeepSeek0006 70.2 DeepSeek V4.1 Flash7 70.3 DeepSeek V4.1 Flash 10178 48.0 DeepSeek V4.1 Flash 10175 79.9 DeepSeek V4.1 Flash 10176 85.3 DeepSeek V4.1 Flash 10179 64.5 DeepSeek V4 Pro 08136 1308 DeepSeek V4.1 Flash
Qwen0007 67.6 Qwen3.8 Max (0902)5 74.1 Qwen3.8 27B6 51.6 Qwen3.8 Max (0902)5 79.9 Qwen3.8 Max (0902)4 86.8 Qwen3.8 Max (0902)6 68.9 Qwen3.7 Max10 1274 Qwen3.8 Max (0902)
xAI0008 67.0 Grok 4.69 66.7 Grok 4.79 44.0 Grok 4.77 78.8 Grok 4.610 79.9 Grok 4.64 74.7 Grok 4.79 1287 Grok 4.6
Z.ai0009 66.9 GLM 5.38 68.5 GLM 5.35 63.8 GLM 5.39 75.7 GLM 5.39 80.5 GLM 5.38 67.6 GLM 5.3 Flash8 1291 GLM 5.3
Xiaomi00010 62.7 MiMo-V2.6-Pro10 65.4 MiMo-V2.6-Pro7 48.2 MiMo-V2.6-Pro11 74.2 MiMo-V2.6-ProNot measuredNot measured7 1307 MiMo-V2.6-Pro
StepFun00011 56.0 Step 5 Preview11 61.2 Step 5 Preview12 29.4 Step 5 Preview12 73.4 Step 5 PreviewNot measuredNot measured17 1195 Step 3.7 Flash
MiniMax00012 54.9 MiniMax M3.1 Flash12 58.9 MiniMax M3.1 Flash10 43.6 MiniMax M3.1 Flash10 74.3 MiniMax M3.1 Flash12 71.2 MiniMax M3Not measured11 1256 MiniMax M3

Best model for your job

The top three for each kind of work, each from the board named on the card. Every best-for list

Today's records

The fastest, the newest, the longest memory, the cheapest working answer, the biggest maker, the hottest download and the best open model, each with the board it was read from. All new and trending models

  • Fastest model 410 tokens / s Nemotron 3.5 Super VL The overall leader, Claude Opus 5.5, runs at 96. 2. 2 Gemini 3.5 Flash Lite 405 3. 3 Claude Haiku 5.5 235 4. 4 Inkling Small 220 5. 5 Muse Spark Open 183 Artificial Analysis, output speed on its own API
  • P Newest release October 28 yesterday P Particle-1 Not ranked yet. 28 models released in the last 30 days. OpenRouter listing dates
  • Longest context 2M tokens Grok 4.20 126 models take a million tokens or more. OpenRouter
  • Cheapest working answer $0.01 per solved task DeepSeek V4 Flash 0731 Solves 88.8% of SWE-bench Verified. Our cost per solved task
  • Most ranked models 25 models OpenAI Of 111 on the LLMBenchmark Index. LLMBenchmark Index
  • Top of Hugging Face trending 116.8K downloads in 30 days DeepSeek V4.1 Flash 1017 Coding index 70.3. Hugging Face
  • Best open weights 70.2 LLMBenchmark Index DeepSeek V4.1 Flash #19 of 111 overall, weights on Hugging Face. LLMBenchmark Index

The state of play, October 2026

points between the top two

3.0

points cover the top ten

10.5

makers in the top ten

3

Anthropic Claude Opus 5.5 leads the overall table with 86.9 on the LLMBenchmark Index, 3.0 points ahead of OpenAI GPT-6 Astra. The top ten are within 10.5 points of each other and come from 3 makers: Anthropic, OpenAI and Google.

Price does not follow rank. Inside the top ten, GPT-6.1 Terra costs $5.00 per million output tokens and scores 76.8, while GPT-6 Astra costs $50.00. For coding work, the cost per solved task tells you more than the token price, because a cheap model that fails costs you the retry.

The best model with public weights, DeepSeek V4.1 Flash, sits at place 19 of 111 with 70.2. Coding has its own leader: Claude Opus 5.5 tops our Vibe Coding Index, which averages six public coding benchmarks and shows how many of them stand behind each score.

28 models were released in the last 30 days, the newest Particle-1 on October 28, 2026. New models often appear here before every board has measured them, so their rows show "No data" until the results are published.

Sources, read today

Every board we read, most models first: how many models it gave us, the bar against the biggest, and when the board itself last changed. Methodology

  1. 351 OpenRouter read October 29, 2026
  2. 218 Epoch AI benchmarking hub read October 29, 2026
  3. 204 Artificial Analysis read October 29, 2026
  4. 188 OpenRouter (provider speed) read October 29, 2026
  5. 188 LMArena (Hugging Face dataset) read October 29, 2026
  6. 156 Hugging Face read October 29, 2026
  7. 147 Design Arena read October 29, 2026
  8. 145 Vals AI boards read October 29, 2026
  9. 132 Artificial Analysis (via OpenRouter) read October 29, 2026
  10. 109 ALE-Bench (Sakana AI) read October 29, 2026
  11. 104 LMArena WebDev read October 29, 2026
  12. 100 MMLU Pro (Vals AI) read October 29, 2026, board updated September 1, 2026
  13. 98 GPQA (Vals AI) read October 29, 2026, board updated September 1, 2026
  14. 73 MathArena read October 29, 2026
  15. 70 Kagi LLM Benchmarking Project read October 29, 2026
  16. 68 Scale SEAL leaderboards read October 29, 2026
  17. 62 AIME (Vals AI) read October 29, 2026, board updated April 16, 2026
  18. 59 Kaggle Benchmarks read October 29, 2026
  19. 43 SWE-bench (swebench.com) read October 29, 2026
  20. 32 EQ-Bench 4 read October 29, 2026
  21. 30 Aider polyglot leaderboard read October 29, 2026
  22. 30 Berkeley Function Calling Leaderboard read October 29, 2026
  23. 28 LiveBench read October 29, 2026, board updated October 25, 2026
  24. 26 Vals Index read October 29, 2026, board updated October 27, 2026
  25. 13 BrowseComp-Plus read October 29, 2026

Questions

What is the best LLM right now?

As of October 29, 2026, Anthropic Claude Opus 5.5 leads 111 models overall with 86.9 ±2.3 on the LLMBenchmark Index. The best model depends on the job, so the tabs above re-rank the table by coding, agents, reasoning, math, knowledge and web design.

Which LLM is best for coding?

Anthropic Claude Opus 5.5 leads our Vibe Coding Index with 88.9, from five benchmarks. 100 models have a coding score; the dots on each row show how many public coding benchmarks stand behind it.

Which LLM is best for agents and tool use?

Anthropic Claude Fable 5.1 leads our Agentic Index with 82.2 as of October 29, 2026. The index averages live benchmarks of tasks where a model works over many steps, in a terminal or a browser, and calls tools.

What is the best cheap LLM?

Among models at or under $1 per million output tokens on OpenRouter, DeepSeek V4.1 Flash has the highest overall index, 70.2, at $0.9085. For coding, the cost per solved task is the better measure of cheap.

What is the best open-weight LLM?

DeepSeek V4.1 Flash has the highest overall index among models with public weights on Hugging Face, 70.2. We link the repository; read its license there.

Where do the numbers come from, and how often are they updated?

From 25 public sources, read every day: OpenRouter for the catalog and prices, the public leaderboards for benchmark results. Every number on the site names its source and the day it was read. We do not run benchmarks ourselves.

Why do some cells say "No data"?

Because that board has not measured that model. We leave the gap visible instead of estimating a number, so every figure you see was published by someone you can check.

T T

Every badge is a model

The top 120 models on the overall index spell the name, number one at the top of the first L. Hover a badge for the model and its place, click it for its page.