Everyday chat
LMArena1
Gemini 4 Carbon
- 2Gemini 4 Argon1520
- 3Claude Opus 5.51509
Read 31 Oct 2026.
Leaderboard · Weekly
Gemini 4 Carbon now leads overall.
Four public leaderboards, read every week and put side by side. Google models lead the human-preference vote and the intelligence index. Anthropic models lead reasoning and coding.
In short
- Gemini 4 Carbon is in the top two on all two boards that list it: #1 on LMArena, #1 on Artificial Analysis.
- New this week: Gemini 4 Carbon entered LMArena at #1 and Gemini 4 Carbon entered Artificial Analysis at #1.
- SWE-Bench Verified, the coding board, has posted no new results since 20 September 2026.
Best for each kind of work
The top three on each board, as the board publishes them.
Reasoning
LiveBench2
Claude Fable 5.1
- 2Claude Opus 5.586.5
- 3GPT-6 Astra86.1
Read 25 Oct 2026.
Coding
SWE-Bench Verified3
Claude 4.5 Opus
- 2Gemini 3 Flash75.8
- 3MiniMax M2.575.8
Read 31 Oct 2026. No new results since 20 September 2026.
Overall intelligence
Artificial Analysis4
Gemini 4 Carbon
- 2Claude Opus 5.558
- 3GPT-6 Astra57
Read 31 Oct 2026. Leader: $8.00 per million tokens, 84.6 tokens a second.
The overall top ten
Models ranked on at least two boards. The score is each model's average standing across its boards, where 100 means top of the board. Our math5
| # | Model | Average standing | Chat | Reasoning | Coding | Index | Change |
|---|---|---|---|---|---|---|---|
| 1 | Gemini 4 CarbonGoogle | #1 | - | - | #1 | New | |
| 2 | Claude Opus 5.5Anthropic | #3 | #2 | - | #2 | ↓ 1 | |
| 3 | GPT-6 AstraOpenAI | - | #3 | - | #3 | ↑ 1 | |
| 4 | Gemini 4 ArgonGoogle | #2 | #5 | - | - | ↓ 1 | |
| 5 | Claude Fable 5.1Anthropic | #5 | #1 | - | #8 | ↓ 3 | |
| 6 | Claude Opus 4.6Anthropic | #7 | - | #4 | - | ||
| 7 | Claude Sonnet 5.5Anthropic | - | #7 | - | #5 | ↑ 4 | |
| 8 | Claude Fable 5Anthropic | #6 | #6 | - | - | ↓ 3 | |
| 9 | Muse WonderMeta | #4 | - | - | #10 | ↓ 2 | |
| 10 | Claude Opus 5Anthropic | #14 | #8 | - | - | ↓ 1 |
Board columns show each model's rank there; a dash means the board does not list it.
What moved this week
- LMArenaGemini 4 Carbon (Google) entered at #1 with 1528.
- LiveBenchNo change in the top ten.
- SWE-Bench VerifiedNo new results since 20 September 2026. The top ten is unchanged.
- Artificial AnalysisGemini 4 Carbon (Google) entered at #1 with 59. GPT-6.1 Terra (OpenAI) entered at #6 with 51. GPT-6 Astra rose to #3 (53 to 57); GPT-6.1 Sol rose to #4 (52 to 55).
Compared with the read of 25 October 2026.
The four boards
Top ten on each, one row per model. Ranks are the board's own; where a board lists a model twice (two effort settings), we keep the better row.
LMArena1
Everyday chat · Human preference rankings from blind A/B comparisons (7M+ votes)
| Rank | Model | Elo | Change |
|---|---|---|---|
| 1 | Gemini 4 Carbon | 1528 | New |
| 2 | Gemini 4 Argon | 1520 | ↓ 1 |
| 3 | Claude Opus 5.5 | 1509 | ↓ 1 |
| 4 | Muse Wonder | 1507 | ↓ 1 |
| 5 | Claude Fable 5.1 | 1504 | |
| 6 | Claude Fable 5 | 1504 | ↓ 2 |
| 7 | Claude Opus 4.6 | 1503 | ↓ 1 |
| 8 | Claude Opus 4.7 | 1501 | ↓ 1 |
| 10 | Gemini 3.8 Flash | 1496 | ↓ 1 |
| 11 | Muse Spark 1.3 | 1495 | ↓ 1 |
Read 31 Oct 2026.
LiveBench2
Reasoning · LiveBench 2.0: contamination-free long-horizon tasks across 7 suites
| Rank | Model | Score | Change |
|---|---|---|---|
| 1 | Claude Fable 5.1 | 88.0 | |
| 2 | Claude Opus 5.5 | 86.5 | |
| 3 | GPT-6 Astra | 86.1 | |
| 4 | GPT-6.1 Sol | 85.4 | |
| 5 | Gemini 4 Argon | 84.1 | |
| 6 | Claude Fable 5 | 83.6 | |
| 7 | Claude Sonnet 5.5 | 82.5 | |
| 8 | Claude Opus 5 | 82.1 | |
| 9 | GPT-6 Sol | 81.9 | |
| 10 | DeepSeek V4.1 Flash 1017 | 79.2 |
Read 25 Oct 2026.
SWE-Bench Verified3
Coding · Real-world software engineering task completion (Verified, 500 tasks)
| Rank | Model | Score | Change |
|---|---|---|---|
| 1 | Claude 4.5 Opus | 76.8 | |
| 2 | Gemini 3 Flash | 75.8 | |
| 3 | MiniMax M2.5 | 75.8 | |
| 4 | Claude Opus 4.6 | 75.6 | |
| 6 | Gemini 3 Pro Preview | 74.2 | |
| 7 | GLM 5 | 72.8 | |
| 8 | GPT 5.2 | 72.8 | |
| 9 | GPT 5.2 Codex | 72.8 | |
| 10 | Claude 4.5 Sonnet | 71.4 | |
| 11 | Kimi K2.5 | 70.8 |
Read 31 Oct 2026. No new results since 20 September 2026.
Artificial Analysis4
Overall intelligence · Artificial Analysis Intelligence Index v5.0 - aggregate of 9 evaluations
| Rank | Model | Index | Change |
|---|---|---|---|
| 1 | Gemini 4 Carbon | 59 | New |
| 2 | Claude Opus 5.5 | 58 | ↓ 1 |
| 3 | GPT-6 Astra | 57 | ↑ 3 |
| 4 | GPT-6.1 Sol | 55 | ↑ 5 |
| 5 | Claude Sonnet 5.5 | 54 | ↓ 3 |
| 6 | GPT-6.1 Terra | 51 | New |
| 7 | Muse Wonder Full | 50 | ↓ 4 |
| 8 | Claude Fable 5.1 | 49 | ↓ 3 |
| 9 | Kimi K3.1 Pro | 46 | ↓ 5 |
| 10 | Muse Wonder | 43 | ↓ 2 |
Read 31 Oct 2026.
Sources
Numbers on the page point here. Reported means the board publishes the figure; Our math means we worked it out from those figures.
- 1LMArena · Human preference rankings from blind A/B comparisons (7M+ votes) · read 31 Oct 2026Reported
- 2LiveBench · LiveBench 2.0: contamination-free long-horizon tasks across 7 suites · read 25 Oct 2026Reported
- 3SWE-Bench Verified · Real-world software engineering task completion (Verified, 500 tasks) · read 31 Oct 2026Reported
- 4Artificial Analysis · Artificial Analysis Intelligence Index v5.0 - aggregate of 9 evaluations · read 31 Oct 2026Reported
- 5Koda overall ranking · average standing across sources 1 to 4 · built 31 Oct 2026Our math
How this page is counted
Once a week we read each board's own published table. One model registry ties a model's different spellings together, so a lab is never counted twice. Boards measure different things, so the overall table averages standings, not raw scores.
Full methodkoda-leaderboard-composite-v1