Benchmarks / Arena (formerly LMArena)

Reported by Arena (formerly LMArena)

Arena (formerly LMArena)

Estimated effect on users praising rather than complaining.

Last updated 2 Oct 2026

Results dated
2 Oct 2026
Results
50 configurations of 48 models
Unit
IPS effect estimate
Licence
Creative Commons Attribution 4.0 International

Praise over complaint: Claude Sonnet 5.5

Top 15 of 48 results · IPS effect estimate, higher is better · lines show the 95% range · ≈ cannot be told apart from the leader. Choose a model to highlight it.Clear highlight

  1. 1≈ GPT-6 Astra (max reasoning)OpenAI 0.35
  2. 2≈ Claude Fable 5.1 (max reasoning)Anthropic 0.32
  3. 3≈ Claude Opus 5.5 (high reasoning)Anthropic 0.31
  4. 4≈ Gemini 4 Argon (high reasoning)Google 0.28
  5. 5≈ Claude Sonnet 5.5 (max reasoning)Anthropic 0.26
  6. 5≈ GPT-6.1 Sol (max reasoning)OpenAI 0.26
  7. 7≈ GPT-6 Sol (max reasoning)OpenAI 0.23
  8. 8 GPT-5.6 Sol (extra-high reasoning)OpenAI 0.20
  9. 9 Claude Fable 5 (high reasoning)Anthropic 0.17
  10. 9 Claude Opus 5 (high reasoning)Anthropic 0.17
  11. 11 Claude Opus 4.8 (high reasoning)Anthropic 0.15
  12. 12 GLM-5.2 (max reasoning)Z.ai 0.10
  13. 13 Claude Sonnet 5 (high reasoning)Anthropic 0.09
  14. 14 GPT-5.5 (extra-high reasoning)OpenAI 0.08
  15. 15 Grok 4.7 (extra-high reasoning)xAI 0.07

Full results

Arena Agent: praise over complaint, IPS effect estimate, higher is better
#ModelPraise over complaint · 95% range
IPS effect estimate, higher is better		Price
$ per million tokens, in / out
1≈ GPT-6 Astra (max reasoning)OpenAI
0.35
0.25–0.44
$10 / $50
2≈ Claude Fable 5.1 (max reasoning)Anthropic
0.32
0.24–0.40
$10 / $50
3≈ Claude Opus 5.5 (high reasoning)Anthropic
0.31
0.23–0.40
$4 / $20
4≈ Gemini 4 Argon (high reasoning)Google
0.28
0.20–0.35
–
5≈ Claude Sonnet 5.5 (max reasoning)Anthropic
0.26
0.13–0.39
$2 / $10
5≈ GPT-6.1 Sol (max reasoning)OpenAI
0.26
0.13–0.38
$2 / $10
7≈ GPT-6 Sol (max reasoning)OpenAI
0.23
0.13–0.33
$2 / $10
8 GPT-5.6 Sol (extra-high reasoning)OpenAI
0.20
0.15–0.24
$4 / $20
9 Claude Fable 5 (high reasoning)Anthropic
0.17
0.13–0.22
$10 / $50
9 Claude Opus 5 (high reasoning)Anthropic · best of 2 settings
0.17
0.11–0.22
$5 / $25
11 Claude Opus 4.8 (high reasoning)Anthropic
0.15
0.10–0.20
$5 / $25
12 GLM-5.2 (max reasoning)Z.ai
0.10
0.07–0.13
$1.40 / $4.40
13 Claude Sonnet 5 (high reasoning)Anthropic
0.09
0.03–0.14
$2 / $10
14 GPT-5.5 (extra-high reasoning)OpenAI · best of 2 settings
0.08
0.04–0.11
$5 / $30
15 Grok 4.7 (extra-high reasoning)xAI
0.07
0.01–0.13
$2 / $6
15 Kimi K3 (max reasoning)Moonshot AI
0.07
0.05–0.09
$3 / $15
15 Hy4 PreviewTencent
0.07
0.04–0.10
$0.83 / $2.50
18 GLM-5.3 (max reasoning)Z.ai
0.05
0.03–0.08
$1.40 / $4.40
18 Gemini 3.8 Flash (high reasoning)Google
0.05
0.02–0.08
$1.50 / $7.50
20 Qwen3.8-Max (0902)Alibaba
0.03
0.01–0.05
$2 / $6
20 Muse Spark 1.3 (max reasoning)Meta
0.03
0.01–0.05
$1.25 / $4.25
20 GPT-5.6 Terra (extra-high reasoning)OpenAI
0.03
-0.01–0.07
$2 / $12
23 DeepSeek-V4.1-Flash (max reasoning)DeepSeek
0.02
0.01–0.04
$0.30 / $1.20
23 DeepSeek-V4-Pro (0813, high reasoning)DeepSeek
0.02
-0.04–0.08
$1.32 / $3.96
25 GPT-5.4 (high reasoning)OpenAI
0.00
-0.03–0.04
$2.50 / $15
25 Grok 4.6 (extra-high reasoning)xAI
0.00
-0.03–0.04
$2 / $6
27 MiMo-V2.6-FlashXiaomi
-0.01
-0.05–0.03
$0.14 / $0.28
27 GLM-5.3-FlashZ.ai
-0.01
-0.02–0.01
$0.15 / $0.50
27 Gemini 3.1 Pro PreviewGoogle
-0.01
-0.04–0.01
$2 / $12
30 GPT-6 Luna (max reasoning)OpenAI
-0.02
-0.05–0.02
$0.10 / $0.50
31 Qwen3.8-27BAlibaba
-0.03
-0.05–-0.00
$0.50 / $3
31 Grok 4.5xAI
-0.03
-0.06–0.00
$2 / $6
31 Gemini 3.7 Flash (high reasoning)Google
-0.03
-0.06–-0.01
$1.50 / $7.50
34 Qwen3.8-Flash-NextAlibaba
-0.04
-0.06–-0.01
–
34 GPT-5.6 Luna (extra-high reasoning)OpenAI
-0.04
-0.06–-0.02
$0.20 / $1.20
36 Step 5 PreviewStepFun
-0.05
-0.10–0.00
–
37 Hy3Tencent
-0.07
-0.10–-0.03
$0.14 / $0.58
38 Gemini 3.6 Flash (high reasoning)Google
-0.09
-0.12–-0.05
$1.50 / $7.50
38 Qwen3.7-MaxAlibaba
-0.09
-0.12–-0.07
$1.48 / $4.42
40 Qwen3.7-PlusAlibaba
-0.11
-0.15–-0.07
$0.32 / $1.28
41 Muse Spark 1.1Meta
-0.12
-0.13–-0.11
$1.25 / $4.25
41 Muse Spark 1.2 (extra-high reasoning)Meta
-0.12
-0.14–-0.10
$1.25 / $4.25
41 MiMo-V2.5-ProXiaomi
-0.12
-0.15–-0.10
$0.43 / $0.87
44 MiniMax-M3MiniMax
-0.14
-0.17–-0.12
$0.30 / $1.20
45 Mistral Medium 3.5Mistral AI
-0.18
-0.21–-0.14
$1.50 / $7.50
46 Inkling SmallThinking Machines
-0.22
-0.26–-0.19
$0.45 / $1.20
47 InklingThinking Machines
-0.23
-0.25–-0.20
$0.95 / $4.05
47 Solar Pro 4Upstage
-0.23
-0.28–-0.19
$0.09 / $0.36

Swipe the table sideways for more columns.

Ranks follow the score as shown, so equal numbers share a rank. ≈ marks results whose 95% range overlaps the leader's: they cannot be told apart from it. Each model is shown at its best setting; show every setting. Results as published by Arena (formerly LMArena); we do not re-run them.

What it measures

Estimated effect on users praising rather than complaining.

What it does not measure

Not a success rate: the estimated effect of choosing this model as the agent's orchestrator, against an even mix of the models tested, inside Arena's own harness.

Source

Arena (formerly LMArena) · Methodology

Contains data from the Arena Leaderboard Dataset by Arena, licensed under CC BY 4.0. Licence: Creative Commons Attribution 4.0 International.