Human preference

LMArena Text Elo leaderboard

21 models have a published score. Gemini 4 Argon (Google) leads with 1520, and the spread to the lowest score, Claude Haiku 5.5 at 1425, is 95 Elo points. The best open-weight result, Kimi K3 at 1488, trails the best closed model, Gemini 4 Argon, by 32. The highest-scoring model you can deploy on OpenLLM Buddy is Kimi K3, ranked 5 of 21 with 1488. Sources report different settings for this test, such as test year, version, or reasoning effort. Each setting is listed in the source table, so compare close scores with care.

Best openllm buddy catalog

1488

Kimi K3

7 models scored

Best open weights

1481

GLM 5.3

5 models scored

Best closed api

1520

Gemini 4 Argon

9 models scored

LMArena Text Elo scores

Elo rating. A 100-point gap means the higher model wins roughly 64% of head-to-head votes.

  • OpenLLM Buddy catalog
  • Open weights
  • Closed API
  1. 1Gemini 4 ArgonLMArena Text, Overall; snapshot Oct 27, 2026; Gemini 4 Argon (High); +/-61520
  2. 2Claude Opus 5.5LMArena Text, Overall; snapshot Oct 27, 2026; Claude Opus 5.5 (High); +/-71509
  3. 3Claude Fable 5.1LMArena Text, Overall; snapshot Oct 27, 2026; Claude Fable 5.1 (Max); +/-61504
  4. 4Gemini 3.8 FlashLMArena Text, Overall; snapshot Oct 27, 2026; Gemini 3.8 Flash (High); +/-41496
  5. 5Kimi K3LMArena text overall, 'Kimi K3 (Max)', ±4, as of 2026-10-271488
  6. 6GLM 5.3GLM-5.3 (Max), ±5, board dated 2026-10-271481
  7. 7GPT-6.1 SolLMArena Text, Overall; snapshot Oct 27, 2026; GPT-6.1 Sol (Max); +/-71481
  8. 8MiMo V2.6 ProLMArena text overall, 'MiMo-V2.6-Pro', ±9, as of 2026-10-271480
  9. 9Claude Sonnet 5.5LMArena Text, Overall; snapshot Oct 27, 2026; Claude Sonnet 5.5 (xHigh); +/-71477
  10. 10GLM 5.3 FlashLMArena text overall, 'GLM-5.3-Flash', ±5, as of 2026-10-271477
  11. 11DeepSeek V4.1 FlashLMArena text overall, 'DeepSeek-V4.1-Flash-20260910 (Max)', ±6, as of 2026-10-271476
  12. 12GPT-6 AstraLMArena Text, Overall; snapshot Oct 27, 2026; GPT-6 Astra (Max); +/-61476
  13. 13DeepSeek V4 ProDeepSeek-V4-Pro (High, 20260813), ±7, board dated 2026-10-271465
  14. 14Gemma 4 31BLMArena text overall, 'Gemma 4 31B', ±7, as of 2026-10-271452
  15. 15Step 5 PreviewStep 5 Preview, ±6, board dated 2026-10-271448
  16. 16Grok 4.7LMArena Text, Overall; snapshot Oct 27, 2026; Grok 4.7 (xHigh); +/-61441
  17. 17Gemma 4 26B A4BLMArena text overall, 'Gemma 4 26B A4B', ±7, as of 2026-10-271438
  18. 18Qwen3.8 27BLMArena text overall, 'Qwen3.8-27B', ±5, as of 2026-10-271436
  19. 19Reflection Beam 1.0Beam 1.0 501B A23B, ±9, board dated 2026-10-271436
  20. 20Nemotron 3 UltraNVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4), ±7, board dated 2026-10-271427
  21. 21Claude Haiku 5.5LMArena Text, Overall; snapshot Oct 27, 2026; Claude Haiku 5.5 (High); +/-71425

What it measures

Visitors chat with two anonymous models and vote for the better answer; votes are turned into an Elo-style rating. It reflects style and helpfulness as much as correctness.

How it is scored

Elo rating. A 100-point gap means the higher model wins roughly 64% of head-to-head votes.

Why it matters

It captures how people experience a model in open-ended conversation, which fixed tests miss.

Official benchmark page

Sources

ModelTestScoreSettingSource
Gemini 4 ArgonLMArena Text Elo1520LMArena Text, Overall; snapshot Oct 27, 2026; Gemini 4 Argon (High); +/-6LMArena (arena.ai) Text leaderboard
Claude Opus 5.5LMArena Text Elo1509LMArena Text, Overall; snapshot Oct 27, 2026; Claude Opus 5.5 (High); +/-7LMArena (arena.ai) Text leaderboard
Claude Fable 5.1LMArena Text Elo1504LMArena Text, Overall; snapshot Oct 27, 2026; Claude Fable 5.1 (Max); +/-6LMArena (arena.ai) Text leaderboard
Gemini 3.8 FlashLMArena Text Elo1496LMArena Text, Overall; snapshot Oct 27, 2026; Gemini 3.8 Flash (High); +/-4LMArena (arena.ai) Text leaderboard
Kimi K3LMArena Text Elo1488LMArena text overall, 'Kimi K3 (Max)', ±4, as of 2026-10-27Arena (LMArena) text leaderboard
GLM 5.3LMArena Text Elo1481GLM-5.3 (Max), ±5, board dated 2026-10-27LMArena (arena.ai) text leaderboard
GPT-6.1 SolLMArena Text Elo1481LMArena Text, Overall; snapshot Oct 27, 2026; GPT-6.1 Sol (Max); +/-7LMArena (arena.ai) Text leaderboard
MiMo V2.6 ProLMArena Text Elo1480LMArena text overall, 'MiMo-V2.6-Pro', ±9, as of 2026-10-27Arena (LMArena) text leaderboard
Claude Sonnet 5.5LMArena Text Elo1477LMArena Text, Overall; snapshot Oct 27, 2026; Claude Sonnet 5.5 (xHigh); +/-7LMArena (arena.ai) Text leaderboard
GLM 5.3 FlashLMArena Text Elo1477LMArena text overall, 'GLM-5.3-Flash', ±5, as of 2026-10-27Arena (LMArena) text leaderboard
DeepSeek V4.1 FlashLMArena Text Elo1476LMArena text overall, 'DeepSeek-V4.1-Flash-20260910 (Max)', ±6, as of 2026-10-27Arena (LMArena) text leaderboard
GPT-6 AstraLMArena Text Elo1476LMArena Text, Overall; snapshot Oct 27, 2026; GPT-6 Astra (Max); +/-6LMArena (arena.ai) Text leaderboard
DeepSeek V4 ProLMArena Text Elo1465DeepSeek-V4-Pro (High, 20260813), ±7, board dated 2026-10-27LMArena (arena.ai) text leaderboard
Gemma 4 31BLMArena Text Elo1452LMArena text overall, 'Gemma 4 31B', ±7, as of 2026-10-27Arena (LMArena) text leaderboard
Step 5 PreviewLMArena Text Elo1448Step 5 Preview, ±6, board dated 2026-10-27LMArena (arena.ai) text leaderboard
Grok 4.7LMArena Text Elo1441LMArena Text, Overall; snapshot Oct 27, 2026; Grok 4.7 (xHigh); +/-6LMArena (arena.ai) Text leaderboard
Gemma 4 26B A4BLMArena Text Elo1438LMArena text overall, 'Gemma 4 26B A4B', ±7, as of 2026-10-27Arena (LMArena) text leaderboard
Qwen3.8 27BLMArena Text Elo1436LMArena text overall, 'Qwen3.8-27B', ±5, as of 2026-10-27Arena (LMArena) text leaderboard
Reflection Beam 1.0LMArena Text Elo1436Beam 1.0 501B A23B, ±9, board dated 2026-10-27LMArena (arena.ai) text leaderboard
Nemotron 3 UltraLMArena Text Elo1427NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4), ±7, board dated 2026-10-27LMArena (arena.ai) text leaderboard
Claude Haiku 5.5LMArena Text Elo1425LMArena Text, Overall; snapshot Oct 27, 2026; Claude Haiku 5.5 (High); +/-7LMArena (arena.ai) Text leaderboard

Scores are published results from model cards, lab reports, and public leaderboards, each linked to its source. Settings such as thinking mode or test version differ between sources and are shown next to each score. OpenLLM Buddy catalog models are uncensored fine-tunes; their scores are those of the official base model. Data collected 2026-10-30. How the index works