Open-Source LLM Leaderboard
As of Nov 10, the top open-weight model on Arena is Kimi K3.1 Pro.
This week
vs Nov 3New models
last 14 daysOpen-Source LLM Leaderboard: full table
Showing 24 rows默认按 LMArena 盲测对决评分排序。新模型(如 Kimi K3.1 Pro)处于初步投票阶段,票数少导致区间较宽。 改按 AA 智能指数排序(查看 K3.1 Pro 综合跑分第一)→
| # | Model | arena.ai Agent Arena | Text↓ | AA index | Price $/1M tokens | Latency (s) & speed | Context | Released | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Net Improvement | Confirmed Success | Praise vs Complaint | Steerability | Intel. | Coding | Input | Output | Lowest route | TTFT | Answer | tok/s | |||||
| 1 | 5.24%±1.13Max | 2.84%±2.44 | 8.74%±3.55 | 7.45%±1.34 | 1488.0±13.0#24 | 46.5 | 64.0 | $2.00 | $10.00 | $2.00 / $10.00 | 2.4 | 74 | 28 | 1.05M | Oct 23 | |
| 2 | 0.57%±0.59Max | 3.47%±1.33 | 0.89%±1.76 | -0.36%±0.85 | 1487.1±4.3#19 | 17.4 | 32.7 | $3.00 | $15.00 | $0.19 / $16.00 ⚠ | – | – | – | 1.05M | Jul 16 | |
| 3 | 0.99%±0.81Max | 2.92%±1.79 | 4.37%±2.45 | -0.95%±1.14 | 1482.5±4.8#26 | 31.7 | 50.5 | $1.40 | $4.40 | $0.07 / $1.40 ⚠ | 1.4 | 30 | 71 | 1.05M | Aug 18 | |
| 4 | 0.40%±0.80Max | 2.05%±1.73 | 1.96%±2.37 | -0.86%±1.19 | 1475.1±4.1#36 | – | 26.1 | $1.40 | $4.40 | $0.01 / $6.80 ⚠ | – | – | – | 1.05M | Jun 16 | |
| 5 | – | – | – | – | 1475.0±6.2#37 | – | 39.3 | $0.30 | $1.20 | $0.006 / $0.50 ⚠ | – | – | – | 1.05M | Sep 10 | |
| 6 | -2.80%±0.71 | 1.68%±1.66 | -4.62%±1.94 | -4.73%±1.11 | 1474.2±4.5#40 | 27.8 | 42.2 | $0.15 | $0.50 | $0.15 / $0.50 | 1.0 | 33 | 62 | 1.05M | Aug 26 | |
| 7 | – | – | – | – | 1466.5±3.6#53 | – | 25.3 | $0.44 | $0.87 | $0.48 / $1.80 | – | – | – | 1.02M | Apr 22 | |
| 8 | – | – | – | – | 1465.2±3.6#56 | – | 23.4 | $1.19 | $3.74 | $1.40 / $4.40 | – | – | – | 205K | Apr 7 | |
| 9 | – | – | – | – | 1464.0±6.7#59 | 23.1 | 32.6 | $1.32 | $3.96 | $0.96 / $1.91 | 1.1 | 28 | 76 | 1.05M | Aug 13 | |
| 10 | – | – | – | – | 1460.8±4.5#61 | – | 26.0 | $0.95 | $4.00 | $0.65 / $3.41 | – | – | – | 262K | Apr 20 | |
| 11 | – | – | – | – | 1457.4±4.3#66 | – | – | $1.00 | $3.20 | $1.00 / $3.20 | – | – | – | 205K | Feb 11 | |
| 12 | -7.67%±1.47 | -12.09%±3.22 | -8.60%±3.91 | -2.59%±1.97 | 1454.2±5.7#76 | – | 24.6 | $0.14 | $0.56 | $0.13 / $0.53 | 1.9 | 24 | 92 | 262K | Jul 6 | |
| 13 | – | – | – | – | 1449.5±3.3#81 | – | – | $0.60 | $3.00 | $0.53 / $3.33 | – | – | – | 262K | Jan 27 | |
| 14 | – | – | – | – | 1441.5±3.1#96 | – | – | $0.60* | $3.60* | $0.60 / $3.60 | – | – | – | 262K | Feb 16 | |
| 15 | -10.42%±1.19 | -17.68%±2.92 | -16.33%±2.67 | -10.34%±2.02 | 1435.9±3.7#104 | 18.9 | 24.6 | $0.30 | $1.20 | $0.30 / $1.20 | 0.7 | 20 | 106 | 1.05M | Jun 1 | |
| 16 | – | – | – | – | 1435.8±4.1#105 | – | 31.2 | $0.14 | $0.28 | $0.091 / $0.18 | – | – | – | 1.05M | Jul 31 | |
| 17 | – | – | – | – | 1433.7±4.2#111 | – | 22.0 | $0.13 | $0.27 | $0.40 / $2.00 | – | – | – | 1M | Apr 22 | |
| 18 | – | – | – | – | 1433.4±4.7#112 | 20.1 | 26.1 | $0.50 | $3.00 | $0.15 / $1.88 ⚠ | 1.3 | 54 | 38 | 1.05M | Aug 14 | |
| 19 | -13.93%±1.80 | -17.71%±4.06 | -18.72%±3.74 | -16.65%±3.00 | 1426.5±6.4#121 | 4.6 | 20.1 | $1.50 | $7.50 | $1.50 / $7.50 | 0.8 | 13 | 168 | 262K | Apr 29 | |
| 20 | – | – | – | – | 1424.7±3.5#125 | – | – | $0.28 | $0.42 | $0.26 / $0.38 | – | – | – | 164K | Dec 1, 2025 | |
| 21 | – | – | – | – | 1424.4±3.9#127 | – | – | $0.58 | $2.20 | $0.50 / $2.00 | – | – | – | 205K | Sep 30, 2025 | |
| 22 | – | – | – | – | 1387.5±5.0#185 | – | – | $0.30* | $1.00* | $0.30 / $1.00 | – | – | – | 262K | Feb 4 | |
| 23 | – | – | – | – | 1302.9±4.1#290 | – | – | $0.48 | $0.50 | $0.36 / $0.40 | – | – | – | 33K | Sep 19, 2024 | |
| 24 | – | – | – | – | 1270.6±8.2#316 | – | – | $0.66* | $1.00* | $0.66 / $1.00 | – | – | – | 33K | Nov 11, 2024 | |
Arena scores are LMArena's Bradley-Terry scale (not Elo). Input/output prices are official API list prices per million tokens; the lowest route price is the cheapest OpenRouter quote. In this view the median Arena margin is ±4.3 while neighbouring models sit a median 3.2 apart (19 of 23 adjacent pairs overlap): gaps under 4.3 say nothing about which model ranks higher.
One row per model: each row uses that model’s best tier for the column you sorted by (best of the tiers that have a measurement, ties going to the stronger tier). Arena scores and prices are model-level, so every tier of a model shares them. Switch to “Each tier” to see each reasoning tier on its own row.
Not yet ranked by Arena 10
Tracked models that have no LMArena or Artificial Analysis score yet, newest first. They move into the table once scored.
GLM-5.3 FlashXZ.AI ·
GPT-6 Astra ProOpenAI ·
Qwen3.8 Flash NextAlibaba ·
Seed 2.1 TurboByteDance ·
Kimi K2.7 CodeMoonshot AI ·
Grok 4.20xAI ·
Cydonia 24B V4.1 (无审查 / 创作破限)TheDrummer ·
Dolphin Mistral 24B (Venice Uncensored / 破限版)Venice ·
Llama 3.1 Euryale 70B v2.2 (无审查 / RP破限)Sao10K ·
Mistral LargeMistral ·
On the Agents board, not yet in our catalogue 20
These models have an arena.ai Agents result but no model page here yet (no price data), so they are not ranked in the table above.
- THy4tencent · 4.66% · arena.ai #14
- XMiMo-V2.6-Proxiaomi · 2.33% · arena.ai #18
- DDeepSeek-V4.1-Flash-20261017 (Max)deepseek · 2.07% · arena.ai #20
- OGPT-6 Luna (Max)openai · 1.52% · arena.ai #23
- THy4 previewtencent · 0.44% · arena.ai #26
- XMiMo-V2.6-Flashxiaomi · 0.09% · arena.ai #30
- TInkling 2 Smallthinky · 0.08% · arena.ai #31
- OGPT-6 Luna (Medium)openai · -1.96% · arena.ai #35
- MMiniMax-M3.1-Flashminimax · -2.76% · arena.ai #36
- SStep 5 Previewstepfun · -3.43% · arena.ai #38
- AQwen3.8-27Balibaba · -4.60% · arena.ai #40
- AQwen3.8-Flash-Nextalibaba · -4.66% · arena.ai #41
- MMistral Large 4mistral · -6.08% · arena.ai #42
- OGPT-6 Luna (Low)openai · -6.23% · arena.ai #43
- SSpark-X2.6synlan · -10.34% · arena.ai #50
- NNemotron 3 Ultranvidia · -14.37% · arena.ai #53
- NNVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4)nvidia · -16.26% · arena.ai #54
- TInklingthinky · -17.07% · arena.ai #55
- TInkling-Smallthinky · -17.68% · arena.ai #56
- USolar Pro 4upstage · -22.84% · arena.ai #57
All 83 model pages
Anthropic 13
OpenAI 13
Alibaba 11
Google 10
Z.AI 7
Meta 6
Moonshot AI 5
DeepSeek 4
xAI 4
Mistral 2
Xiaomi 2
ByteDance 1
MiniMax 1
Sao10K 1
Tencent 1
TheDrummer 1
Venice 1
Sources & methodology
As of Nov 10, 2026, the highest-rated open-weight model on Arena is Kimi K3.1 Pro (1,488.0), followed by Kimi K3 (1,487.1) and GLM-5.3 (1,482.5). 24 open-weight models can be self-hosted; open weights do not mean unrestricted commercial use, so check each licence.
- AGarena.ai Agent Arena
Agent Arena (Capabilities · Overall): Net Improvement, Confirmed Success, Praise vs Complaint and Steerability with 95% CIs, sessions and ranks (arena.ai, via the public Hugging Face dataset lmarena-ai/leaderboard-dataset, configs agent / agent_task_outcome_explicit / agent_praise_complaint / agent_steerability, CC BY 4.0). Hugging Face
- LMLMArena
Arena score, 95% confidence interval, votes and rank (overall Text leaderboard with style control, the same view arena.ai shows by default; dataset CC BY 4.0).
- AAArtificial Analysis
Intelligence Index, Coding Index, output speed, TTFT / first-answer latency, release dates and official prices (shown on this page only, not in the public API).
- OROpenRouter
Per-provider prices (the cheapest becomes “lowest route”), context length and listing dates.
- $Vendor pricing pages
“Input / Output” are official API list prices (USD per 1M tokens); when no official price exists we use the OpenRouter price and mark it with *.
How we compute it
- Ranking: by default identical to arena.ai Agent Arena (Capabilities · Overall) — sorted by Net Improvement, using arena's own rank. Rank spread comes from 95% CIs: best = 1 + models whose lower bound exceeds this model's upper bound; worst = 1 + other models whose upper bound exceeds this model's lower bound. Models missing from the public dataset (mostly ones arena marks Preliminary) are hand-entered from the arena.ai page with the entry date — only values visible on the page, nothing estimated; once the dataset includes them it takes over, and entries older than 21 days raise a warning. Models without agent results follow, sorted by Text Arena score (style control; "#n" is the text rank).
- Rank: by default, LMArena Text Arena score with style control (the same view arena.ai shows by default; Bradley-Terry, not Elo). The ± is the half-width of the 95% confidence interval, and neighbouring ranks often overlap. Models with fewer than 5,000 votes or ± above 10 are marked Preliminary. We do not invent a composite score.
- New models: models released within 30 days (up to 6) that have an AA Intelligence Index but no Arena Text or Agents score are shown in a “New · no Arena score yet” group inside the top 30 rows of the default view, ordered by AA Intelligence Index and marked “New” without taking a rank; they move into the ranking automatically once Arena lists them. Models without even an AA index are listed under “Not ranked yet” at the bottom. Vendor names are merged (e.g. Kimi → Moonshot AI); open-weight status comes from upstream metadata, gaps are filled only from a sourced override list, and unknowns show –.
- Intelligence / Coding Index: Artificial Analysis aggregates of public benchmarks. When a model has several reasoning-effort tiers, the table shows the best tier for the current sort column; switch to “by tier” to see each one.
- Prices are USD per 1M tokens. “Lowest route” is the cheapest OpenRouter provider’s input / output price; ⚠ marks routes more than 60% below list (usually promos or a single provider). The best-value board only includes models with Arena ≥ 1400 or Intelligence Index ≥ 40, sorted by official output price.
- Latency and speed: TTFT is the time to the first token (including thinking tokens); “Answer” is the time to the first answer token (includes the thinking phase for reasoning models); speed is output tokens per second. All are Artificial Analysis medians.
- Updates: Text and other data sync daily at 14:20 (UTC+8), the Agents board at 14:25. “This week” compares against the daily record from 7 days earlier (or the oldest one, with its age shown, while fewer than 7 days exist); rank arrows in the table appear after 14 days of records.
- Updates: synced automatically every day at 14:20 (UTC+8). LMArena publishes on its own schedule, so the “Arena published” date can be older than the update date. “This week” compares against the daily archive from 7 days earlier.
- Open data: /api/public/llm-leaderboard.php (JSON, LMArena and OpenRouter fields only).
FAQ
What is the best AI model right now?
As of Nov 10, 2026, the arena.ai Agents leaderboard (Net Improvement in real agent sessions) is led by Claude Opus 5.5 (19.96%, 95% interval 17.90%–22.01%), ahead of Gemini 4 Carbon (16.63%). The top intervals overlap heavily, so the exact order is not reliable. On the LMArena Text leaderboard, Gemini 4 Carbon leads (1,526.0).
Which good AI model is the cheapest?
Among models above the quality floor (Arena ≥ 1400 or Intelligence Index ≥ 40), MiMo V2.5 has the lowest official price: $0.13 input / $0.27 output per 1M tokens, followed by DeepSeek V4 Flash ($0.28 output). See the best-value leaderboard.
What is the best open-source (open-weight) AI model?
By Arena score, the top open-weight model is Kimi K3.1 Pro (Moonshot AI, 1,488.0), followed by Kimi K3 (1,487.1). We track 24 open-weight models with an Arena score — see the open-weight leaderboard. Open weights do not mean unrestricted commercial use; check each licence.
Which AI model is best for coding?
On the Artificial Analysis Coding Index, Gemini 4 Carbon is highest (65.2), followed by Kimi K3.1 Pro (64.0). The index aggregates coding benchmarks and differs from the Arena score; see the full coding leaderboard.
Which AI model is the fastest?
The fastest measured output speed belongs to Claude Haiku 5.5 (234 tokens/s); the lowest time to first answer token belongs to Mistral Medium 3.5 (12.71 s). These are Artificial Analysis medians; for reasoning models the first-answer latency includes thinking time.
Which AI models were released recently?
2 models released in the last 14 days are listed, including: Claude Fable 5.2 (Nov 9, 2026), Gemini 4 Carbon (Oct 28, 2026). New models usually take days to weeks to enter the LMArena ranking; until then, any model with an Artificial Analysis Intelligence Index is shown in the “New · no Arena score yet” group within the top 30 rows of the default view, ordered by that index and marked “New” without taking a rank.
How often is the data updated, and where does it come from?
Synced automatically every day: the Agents board comes from arena.ai (public Hugging Face dataset, current release Nov 8, 2026), Text scores from LMArena (released Nov 9, 2026); intelligence and coding indices, speed and latency from Artificial Analysis; route prices and context from OpenRouter. This page was updated Nov 10, 2026.
Why do prices differ from the official site?
Input/output prices are official list prices in USD per million tokens. The lowest route price is the cheapest OpenRouter route, which may be a promotion or a single provider and can differ from local resellers.