AI model rankings
Which models lead each category on independent, third-party leaderboards. Every number links to the leaderboard it came from and shows the date we checked it.
Writing & chat
Human-preference votes on everyday prompts.
LMArena Text Arena (overall)
Arena score (Bradley-Terry, style control) ± 95% CI · Source: lmarena.ai · leaderboard date 2026-10-15 · checked 2026-10-15
| Rank on source | Model | Provider | Score | Benchmark | Notes |
|---|---|---|---|---|---|
| 1 | gemini-4-argon-high | 1524±9 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | Preliminary; 4,904 votes. Rank is the displayed rank; rank spread shown on source. | |
| 2 | muse-wonder-max | Meta | 1509±11 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 3,247 votes. Rank is the displayed rank; rank spread shown on source. |
| 3 | claude-opus-5.5-high | Anthropic | 1508±7 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 8,300 votes. Rank is the displayed rank; rank spread shown on source. |
| 4 | claude-fable-5-high | Anthropic | 1504±4 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 42,004 votes. Rank is the displayed rank; rank spread shown on source. |
| 5 | claude-opus-4-6-high | Anthropic | 1503±3 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 81,394 votes. Rank is the displayed rank; rank spread shown on source. |
| 6 | claude-fable-5.1-max | Anthropic | 1502±6 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 14,590 votes. Rank is the displayed rank; rank spread shown on source. |
| 9 | gemini-3.8-flash-high | 1496±5 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | Preliminary; 37,454 votes. Rank is the displayed rank; rank spread shown on source. | |
| 10 | muse-spark-1.3-max | Meta | 1494±6 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 17,561 votes. Rank is the displayed rank; rank spread shown on source. |
| 18 | kimi-k3-max | Moonshot | 1487±5 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 32,795 votes. Rank is the displayed rank; rank spread shown on source. |
| 22 | gpt-6.1-sol-max | OpenAI | 1484±8 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 6,320 votes; top OpenAI entry. Rank is the displayed rank; rank spread shown on source. |
| 27 | mimo-v2.6-pro | Xiaomi | 1479±9 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 4,075 votes; open weights (MIT). Rank is the displayed rank; rank spread shown on source. |
| 28 | glm-5.3-max | Z.ai | 1479±5 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 22,543 votes; open weights (MIT). Rank is the displayed rank; rank spread shown on source. |
| 30 | gpt-6-astra-max | OpenAI | 1477±6 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 12,163 votes. Rank is the displayed rank; rank spread shown on source. |
| 36 | claude-sonnet-5.5-xhigh | Anthropic | 1475±8 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 6,831 votes. Rank is the displayed rank; rank spread shown on source. |
| 38 | deepseek-v4.1-flash-max | DeepSeek | 1475±6 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 12,793 votes; top DeepSeek entry. Rank is the displayed rank; rank spread shown on source. |
| 52 | grok-4.5 | xAI (SpaceXAI) | 1466±4 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 39,896 votes; top xAI entry. Rank is the displayed rank; rank spread shown on source. |
| 89 | grok-4.7-xhigh | xAI (SpaceXAI) | 1444±6 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 11,591 votes. Rank is the displayed rank; rank spread shown on source. |
| 91 | gpt-6-luna-max | OpenAI | 1443±5 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 49,617 votes. Rank is the displayed rank; rank spread shown on source. |
| 111 | mistral-large-4 | Mistral | 1432±7 | LMArena Text Arena (overall) lmarena.ai, 2026-10-15 | 7,451 votes; top Mistral entry. Rank is the displayed rank; rank spread shown on source. |
Reasoning & general capability
Hard reasoning tests and the sources' own capability indexes.
Epoch Capabilities Index (ECI)
ECI (Epoch composite of 50+ benchmarks) · Source: epoch.ai · leaderboard date 2026-10-15 · checked 2026-10-15
| Rank on source | Model | Provider | Score | Benchmark | Notes |
|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | Anthropic | 166.82 | Epoch Capabilities Index (ECI) epoch.ai, 2026-10-15 | CI 163.54-170.99; model date 2026-09-22. ECI is Epoch's own published index; we did not combine anything. Rank by sorting Epoch's CSV. |
| 2 | GPT-6 Astra | OpenAI | 166.41 | Epoch Capabilities Index (ECI) epoch.ai, 2026-10-15 | CI 163.35-171.23; model date 2026-09-03. ECI is Epoch's own published index; we did not combine anything. Rank by sorting Epoch's CSV. |
| 3 | GPT-6.1 Sol | OpenAI | 166.32 | Epoch Capabilities Index (ECI) epoch.ai, 2026-10-15 | CI 163.17-170.82; model date 2026-09-29. ECI is Epoch's own published index; we did not combine anything. Rank by sorting Epoch's CSV. |
| 4 | Claude Fable 5.1 | Anthropic | 164.65 | Epoch Capabilities Index (ECI) epoch.ai, 2026-10-15 | CI 161.92-168.48; model date 2026-09-01. ECI is Epoch's own published index; we did not combine anything. Rank by sorting Epoch's CSV. |
| 5 | Gemini 4 Argon | 164.56 | Epoch Capabilities Index (ECI) epoch.ai, 2026-10-15 | CI 161.67-168.7; model date 2026-09-30. ECI is Epoch's own published index; we did not combine anything. Rank by sorting Epoch's CSV. | |
| 6 | Claude Sonnet 5.5 | Anthropic | 164.45 | Epoch Capabilities Index (ECI) epoch.ai, 2026-10-15 | CI 161.35-168.26; model date 2026-09-28. ECI is Epoch's own published index; we did not combine anything. Rank by sorting Epoch's CSV. |
| 7 | Claude Opus 5 | Anthropic | 162.6 | Epoch Capabilities Index (ECI) epoch.ai, 2026-10-15 | CI 160.0-165.75; model date 2026-07-24. ECI is Epoch's own published index; we did not combine anything. Rank by sorting Epoch's CSV. |
| 8 | GPT-6 Sol | OpenAI | 162.55 | Epoch Capabilities Index (ECI) epoch.ai, 2026-10-15 | CI 160.03-165.84; model date 2026-09-22. ECI is Epoch's own published index; we did not combine anything. Rank by sorting Epoch's CSV. |
ARC-AGI-2 (semi-private)
% tasks solved · Source: arcprize.org · checked 2026-10-15
| Rank on source | Model | Provider | Score | Benchmark | Notes |
|---|---|---|---|---|---|
| – | GPT-6 Astra (Max) | OpenAI | 95.0 | ARC-AGI-2 (semi-private) arcprize.org, 2026-10-15 | Cost/task US$1.12. Model date 2026-09-02. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. |
| – | GPT-6.1 Sol (Max) | OpenAI | 94.2 | ARC-AGI-2 (semi-private) arcprize.org, 2026-10-15 | Cost/task US$0.25. Model date 2026-09-29. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. |
| – | GPT-6 Astra (XHigh) | OpenAI | 93.3 | ARC-AGI-2 (semi-private) arcprize.org, 2026-10-15 | Cost/task US$0.83. Model date 2026-09-02. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. |
| – | Claude Opus 5.5 (High) | Anthropic | 93.3 | ARC-AGI-2 (semi-private) arcprize.org, 2026-10-15 | Cost/task US$0.41. Model date 2026-09-22. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. |
| – | GPT-5.6 Sol (Max) | OpenAI | 92.5 | ARC-AGI-2 (semi-private) arcprize.org, 2026-10-15 | Cost/task US$1.44. Model date 2026-07-09. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. |
| – | Claude Opus 5.5 (XHigh) | Anthropic | 92.5 | ARC-AGI-2 (semi-private) arcprize.org, 2026-10-15 | Cost/task US$0.67. Model date 2026-09-22. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. |
| – | GPT-6 Astra (High) | OpenAI | 92.1 | ARC-AGI-2 (semi-private) arcprize.org, 2026-10-15 | Cost/task US$0.67. Model date 2026-09-02. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. |
| – | GPT-6 Astra (Medium) | OpenAI | 92.1 | ARC-AGI-2 (semi-private) arcprize.org, 2026-10-15 | Cost/task US$0.48. Model date 2026-09-02. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. |
| – | Claude Opus 5.5 (Max) | Anthropic | 91.7 | ARC-AGI-2 (semi-private) arcprize.org, 2026-10-15 | Cost/task US$1.85. Model date 2026-09-22. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. |
| – | GPT-6.1 Sol (XHigh) | OpenAI | 91.7 | ARC-AGI-2 (semi-private) arcprize.org, 2026-10-15 | Cost/task US$0.18. Model date 2026-09-29. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. |
ARC-AGI-3 (semi-private)
% tasks solved · Source: arcprize.org · checked 2026-10-15
| Rank on source | Model | Provider | Score | Benchmark | Notes |
|---|---|---|---|---|---|
| – | GPT-6 Astra - Provider Adapter (High) | OpenAI | 99.9 | ARC-AGI-3 (semi-private) arcprize.org, 2026-10-15 | Model date 2026-09-02. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. ARC-AGI-3 uses the Provider Adapter harness. |
| – | GPT-6 Astra - Provider Adapter (Max) | OpenAI | 98.6 | ARC-AGI-3 (semi-private) arcprize.org, 2026-10-15 | Model date 2026-09-02. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. ARC-AGI-3 uses the Provider Adapter harness. |
| – | GPT-6 Astra - Provider Adapter (XHigh) | OpenAI | 98.4 | ARC-AGI-3 (semi-private) arcprize.org, 2026-10-15 | Model date 2026-09-02. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. ARC-AGI-3 uses the Provider Adapter harness. |
| – | GPT-6 Astra - Provider Adapter (Medium) | OpenAI | 98.4 | ARC-AGI-3 (semi-private) arcprize.org, 2026-10-15 | Model date 2026-09-02. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. ARC-AGI-3 uses the Provider Adapter harness. |
| – | GPT-6 Astra - Provider Adapter (Low) | OpenAI | 98.0 | ARC-AGI-3 (semi-private) arcprize.org, 2026-10-15 | Model date 2026-09-02. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. ARC-AGI-3 uses the Provider Adapter harness. |
| – | Gemini 4 Argon - Provider Adapter (Low) | 97.5 | ARC-AGI-3 (semi-private) arcprize.org, 2026-10-15 | Model date 2026-10-13. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. ARC-AGI-3 uses the Provider Adapter harness. | |
| – | Gemini 4 Argon - Provider Adapter (Medium) | 97.3 | ARC-AGI-3 (semi-private) arcprize.org, 2026-10-15 | Model date 2026-10-13. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. ARC-AGI-3 uses the Provider Adapter harness. | |
| – | Gemini 4 Argon - Provider Adapter (High) | 97.3 | ARC-AGI-3 (semi-private) arcprize.org, 2026-10-15 | Model date 2026-10-13. Read from arcprize.org's leaderboard data file; leaderboard is a chart with no ranks or dates. ARC-AGI-3 uses the Provider Adapter harness. |
LiveBench 2.0 Reasoning
Suite score (0-100) · Source: livebench.ai · leaderboard date 2026-10-14 · checked 2026-10-15
| Rank on source | Model | Provider | Score | Benchmark | Notes |
|---|---|---|---|---|---|
| 1 | Gemini 4 Argon | 95.1 | LiveBench 2.0 Reasoning livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). | |
| 2 | GPT-6 Astra | OpenAI | 92.7 | LiveBench 2.0 Reasoning livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 3 | GPT-6.1 Sol | OpenAI | 92.6 | LiveBench 2.0 Reasoning livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 4 | Claude Opus 5.5 | Anthropic | 92.2 | LiveBench 2.0 Reasoning livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 5 | Claude Fable 5.1 | Anthropic | 91.7 | LiveBench 2.0 Reasoning livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 6 | GPT-5.6 Sol | OpenAI | 91.7 | LiveBench 2.0 Reasoning livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 7 | Claude Sonnet 5.5 | Anthropic | 91.6 | LiveBench 2.0 Reasoning livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 8 | Claude Opus 5 | Anthropic | 91.2 | LiveBench 2.0 Reasoning livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 9 | Kimi K3 | Moonshot | 90.7 | LiveBench 2.0 Reasoning livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 10 | Muse Wonder 1.0 | Meta | 90.6 | LiveBench 2.0 Reasoning livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 11 | GPT-5.6 Terra | OpenAI | 90.6 | LiveBench 2.0 Reasoning livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 12 | Grok 4.6 | xAI (SpaceXAI) | 90.5 | LiveBench 2.0 Reasoning livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
LiveBench 2.0 Overall
Mean of suite scores (LiveBench's own) · Source: livebench.ai · leaderboard date 2026-10-14 · checked 2026-10-15
| Rank on source | Model | Provider | Score | Benchmark | Notes |
|---|---|---|---|---|---|
| 1 | Claude Fable 5.1 | Anthropic | 88.0 | LiveBench 2.0 Overall livebench.ai, 2026-10-15 | LiveBench's own overall; listed order. |
| 2 | Claude Opus 5.5 | Anthropic | 86.5 | LiveBench 2.0 Overall livebench.ai, 2026-10-15 | LiveBench's own overall; listed order. |
| 3 | GPT-6 Astra | OpenAI | 86.1 | LiveBench 2.0 Overall livebench.ai, 2026-10-15 | LiveBench's own overall; listed order. |
| 4 | GPT-6.1 Sol | OpenAI | 85.4 | LiveBench 2.0 Overall livebench.ai, 2026-10-15 | LiveBench's own overall; listed order. |
| 5 | Gemini 4 Argon | 84.1 | LiveBench 2.0 Overall livebench.ai, 2026-10-15 | LiveBench's own overall; listed order. | |
| 6 | Claude Fable 5 | Anthropic | 83.6 | LiveBench 2.0 Overall livebench.ai, 2026-10-15 | LiveBench's own overall; listed order. |
| 7 | Claude Sonnet 5.5 | Anthropic | 82.5 | LiveBench 2.0 Overall livebench.ai, 2026-10-15 | LiveBench's own overall; listed order. |
| 8 | Claude Opus 5 | Anthropic | 82.1 | LiveBench 2.0 Overall livebench.ai, 2026-10-15 | LiveBench's own overall; listed order. |
| 9 | GPT-6 Sol | OpenAI | 81.9 | LiveBench 2.0 Overall livebench.ai, 2026-10-15 | LiveBench's own overall; listed order. |
| 10 | Muse Wonder 1.0 | Meta | 78.1 | LiveBench 2.0 Overall livebench.ai, 2026-10-15 | LiveBench's own overall; listed order. |
| 11 | DeepSeek V4.1 Flash | DeepSeek | 77.3 | LiveBench 2.0 Overall livebench.ai, 2026-10-15 | LiveBench's own overall; listed order. |
| 12 | Claude Opus 4.8 | Anthropic | 76.8 | LiveBench 2.0 Overall livebench.ai, 2026-10-15 | LiveBench's own overall; listed order. |
Coding
Real software tasks.
LiveBench 2.0 Coding
Suite score (0-100) · Source: livebench.ai · leaderboard date 2026-10-14 · checked 2026-10-15
| Rank on source | Model | Provider | Score | Benchmark | Notes |
|---|---|---|---|---|---|
| 1 | Claude Fable 5 | Anthropic | 98.3 | LiveBench 2.0 Coding livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 2 | Claude Fable 5.1 | Anthropic | 97.7 | LiveBench 2.0 Coding livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 3 | Claude Opus 5.5 | Anthropic | 96.3 | LiveBench 2.0 Coding livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 4 | GPT-6.1 Sol | OpenAI | 96.0 | LiveBench 2.0 Coding livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 5 | Claude Sonnet 5.5 | Anthropic | 94.7 | LiveBench 2.0 Coding livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 6 | Grok 4.5 | xAI (SpaceXAI) | 94.3 | LiveBench 2.0 Coding livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 7 | GPT-6 Astra | OpenAI | 93.7 | LiveBench 2.0 Coding livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 8 | GPT-6 Sol | OpenAI | 93.1 | LiveBench 2.0 Coding livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 9 | Claude Opus 5 | Anthropic | 90.8 | LiveBench 2.0 Coding livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 10 | Muse Spark 1.3 | Meta | 90.2 | LiveBench 2.0 Coding livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
| 11 | Gemini 4 Argon | 89.7 | LiveBench 2.0 Coding livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). | |
| 12 | GPT-5.6 Sol | OpenAI | 89.7 | LiveBench 2.0 Coding livebench.ai, 2026-10-15 | Rank computed by sorting the source's sortable column (ties: source order). |
SWE-bench Pro (public set)
% resolved · Source: scale.com · checked 2026-10-15
| Rank on source | Model | Provider | Score | Benchmark | Notes |
|---|---|---|---|---|---|
| 1 | Muse Spark 1.1 | Meta | 61.50±3.10 | SWE-bench Pro (public set) scale.com, 2026-10-15 | No date on page; newest listed models predate current flagships, so treat as possibly stale. |
| 2 | gpt-5.4 (xHigh) | OpenAI | 59.10±3.56 | SWE-bench Pro (public set) scale.com, 2026-10-15 | No date on page; newest listed models predate current flagships, so treat as possibly stale. |
| 3 | Muse Spark | Meta | 55.00±3.60 | SWE-bench Pro (public set) scale.com, 2026-10-15 | No date on page; newest listed models predate current flagships, so treat as possibly stale. |
| 4 | claude-opus-4-6 (thinking) | Anthropic | 51.90±3.61 | SWE-bench Pro (public set) scale.com, 2026-10-15 | No date on page; newest listed models predate current flagships, so treat as possibly stale. |
| 5 | gemini-3.1-pro (thinking) | 46.10±3.60 | SWE-bench Pro (public set) scale.com, 2026-10-15 | No date on page; newest listed models predate current flagships, so treat as possibly stale. |
Price-performance
What it costs a source to run its tasks. Lower is cheaper.
LiveBench 2.0 cost per successful task
USD per successful task · Source: livebench.ai · leaderboard date 2026-10-14 · checked 2026-10-15
| Rank on source | Model | Provider | Score | Benchmark | Notes |
|---|---|---|---|---|---|
| – | DeepSeek V4.1 Flash Lowest cost | DeepSeek | 0.134 | LiveBench 2.0 cost per successful task livebench.ai, 2026-10-15 | Shown for models with LiveBench overall >= 70 (a display filter, not a blend). |
| – | GLM-5.3 Flash | Z.ai | 0.149 | LiveBench 2.0 cost per successful task livebench.ai, 2026-10-15 | Shown for models with LiveBench overall >= 70 (a display filter, not a blend). |
| – | Claude Haiku 5.5 | Anthropic | 0.211 | LiveBench 2.0 cost per successful task livebench.ai, 2026-10-15 | Shown for models with LiveBench overall >= 70 (a display filter, not a blend). |
| – | GPT-6.1 Sol | OpenAI | 0.734 | LiveBench 2.0 cost per successful task livebench.ai, 2026-10-15 | Shown for models with LiveBench overall >= 70 (a display filter, not a blend). |
| – | Grok 4.5 | xAI (SpaceXAI) | 0.74 | LiveBench 2.0 cost per successful task livebench.ai, 2026-10-15 | Shown for models with LiveBench overall >= 70 (a display filter, not a blend). |
| – | Muse Spark 1.3 | Meta | 0.974 | LiveBench 2.0 cost per successful task livebench.ai, 2026-10-15 | Shown for models with LiveBench overall >= 70 (a display filter, not a blend). |