OverviewAll LeaderboardsMethodology
New Chat

Work Index

Which model is the best at knowledge work?

Nov 18, 2026
1,819,878 sessions
68 models

Work Index

Frontier
AnthropicClaude Fable 5.2 (Max)1,445 on the index. 128 points above rank 2.
Daily Driver
GPT-6.1 Terra (Max)Near-frontier performance at 1/10th the price.
Efficient
MetaMuse Ember (Max)Solid performance at 1/100th the price.
1,1001,2001,3001,4001,500
  1. 1,445
    Anthropic
    Claude Fable 5.2Max
  2. 1,317
    Anthropic
    Claude Opus 5.5Max
  3. 1,306
    Gemini 4 NeonHighPreliminary
  4. 1,288
    Anthropic
    Claude Sonnet 5.5Max
  5. 1,285
    Gemini 4 CarbonHigh
  6. 1,280
    GPT-6.2 AstraMax
  7. 1,275
    Anthropic
    Claude Fable 5.1Max
  8. 1,269
    Anthropic
    Claude Opus 5.5High
  9. 1,267
    GPT-6 AstraMax
  10. 1,252
    GPT-6.2 SolMax
  11. 1,235
    Gemini 4 HydrogenHighPreliminary
  12. 1,233
    GPT-6.1 SolMax
  13. 1,232
    GPT-6.1 TerraMax
  14. 1,215
    Gemini 4 ArgonHigh
  15. 1,199
    GPT-6.1 SolMedium
Model RankingIndexCost per task (P50)Time per task (P50)
  1. 1AnthropicClaude Fable 5.2 (Max)1,445$11.3225.2 min
  2. 2AnthropicClaude Opus 5.5 (Max)1,317$6.9122 min
  3. 3Gemini 4 Neon (High)Preliminary1,306$2.969.6 min
  4. 4AnthropicClaude Sonnet 5.5 (Max)1,288$4.2420.4 min
  5. 5Gemini 4 Carbon (High)1,285$1.6211.5 min
  6. 6GPT-6.2 Astra (Max)1,280$4.1713.8 min
  7. 7AnthropicClaude Fable 5.1 (Max)1,275$10.7518.8 min
  8. 8AnthropicClaude Opus 5.5 (High)1,269$2.646.6 min
  9. 9GPT-6 Astra (Max)1,267$7.5013.4 min
  10. 10GPT-6.2 Sol (Max)1,252$1.2210.2 min
  11. 11Gemini 4 Hydrogen (High)Preliminary1,235$1.806.1 min
  12. 12GPT-6.1 Sol (Max)1,233$1.3313.6 min
  13. 13GPT-6.1 Terra (Max)1,232$1.0414.8 min
  14. 14Gemini 4 Argon (High)1,215$0.899.7 min
  15. 15GPT-6.1 Sol (Medium)1,199$0.474 min

Pareto frontier

Best index for the cost
8009001,0001,1001,2001,3001,400
MetaAnthropicAnthropic
$0.01$0.10$1$10

Median cost per task (last 14 days)

Frontier
AnthropicClaude Fable 5.2 (Max)1,445 on the index. 128 points above rank 2.
Daily Driver
GPT-6.1 Terra (Max)Near-frontier performance at 1/10th the price.
Efficient
MetaMuse Ember (Max)Solid performance at 1/100th the price.
On the frontierIndex$/task
  1. AnthropicClaude Fable 5.2 (Max)1,445$11.32
  2. AnthropicClaude Opus 5.5 (Max)1,317$6.91
  3. Gemini 4 Neon (High)1,306$2.96
  4. Gemini 4 Carbon (High)1,285$1.62
  5. GPT-6.2 Sol (Max)1,252$1.22
  6. GPT-6.1 Terra (Max)1,232$1.04
  7. Gemini 4 Argon (High)1,215$0.89
  8. GPT-6.1 Sol (Medium)1,199$0.47
  9. GPT-6.1 Terra (Medium)1,160$0.15
  10. DeepSeek-V4.1-Flash-20261017 (Max)1,112$0.14
  11. MetaMuse Ember (Max)1,110$0.02
  12. GPT-6 Luna (Low)1,018$0.01
8 models not plotted

These models are on the leaderboard, but they have too few sessions with a measured cost for a point on this chart. The numbers beside their names show their ranks.

  • 17Muse Wonder Full (Max)
  • 24Kimi K3.1 Pro (Max)
  • 53Beam 1.0 501B A23B
  • 58Muse Spark 1.1
  • 63Mistral Medium 3.5
  • 64NVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4)
  • 66Inkling
  • 67Inkling-Small

Signal Performance

Ranking by Task CompletionScore
  1. 1AnthropicClaude Sonnet 5.5 (Max)1,307
  2. 2AnthropicClaude Fable 5.2 (Max)1,304
  3. 3Gemini 4 Neon (High)1,272
  4. 4AnthropicClaude Opus 5.5 (Max)1,267
  5. 5Gemini 4 Carbon (High)1,255
  6. 6GPT-6 Astra (Max)1,243
  7. 7AnthropicClaude Fable 5.1 (Max)1,228
  8. 8GPT-6.2 Sol (Max)1,227
  9. 9Gemini 4 Hydrogen (High)1,216
  10. 10Gemini 4 Argon (High)1,210
  11. 11GPT-6.1 Sol (Max)1,209
  12. 12GPT-6.2 Astra (Max)1,207
  13. 13MetaMuse Wonder Full (Max)1,203
  14. 14AnthropicClaude Opus 5.5 (High)1,199
  15. 15AnthropicClaude Opus 5 (Max)1,197

Each rank counts all 68 models

Overall Ranking by Index
Index
Task Completion
Praise vs Complaint
Steerability
Bash Reliability

Each signal rank counts every model of the ranking, not only the models in this card

Previous
Alignment Index

Which agent best aligns with your intent and reports honestly?

Next
Coding Index

Which agents are best at real coding work?

Frequently asked questions

Agent Mode

Try Agent Mode

Put these models to work on your own real tasks in Agent Mode.

Get started
How the Agent Leaderboard works

How the Agent Leaderboard works

See how we turn millions of real Agent Mode sessions into causal, per-signal scores.

Read the methodology

USE CASES

  • Chat with AI
  • Build Apps & Websites
  • Write & Edit Text
  • Search the Web
  • Generate Images
  • Generate Videos
  • Chose any model
  • Compare Models Side by Side

LEADERBOARD RANKINGS

  • Overall
  • Agent
  • Alignment Index
  • Text
  • WebDev
  • Image-to-WebDev
  • Text to Image
  • Image Edit
  • Text to Video
  • Image to Video
  • Video Edit
  • Vision
  • Document
  • Search

COMPANY

  • About Us
  • How It Works
  • Blog
  • Careers
  • Leaderboard Changelog
  • Product Changelog
  • Help Center
  • FAQ

LEGAL

  • Terms
  • Privacy
  • Cookies

FOLLOW

  • X
  • LinkedIn
  • YouTube
  • Discord

© Arena Intelligence