AI Model & Agent Leaderboards
Which model or agent should you use for the job? We collect results from public evaluations and official pricing pages, grouped by use case. Every figure links to where it was published and shows the date it applies to. We never estimate or fill in missing numbers.
Last updated
Coding
Which models write and fix real code best?
Code Arena (WebDev): top 8
People compare two anonymous models on the same web-development task and vote for the better result. Higher is better; overlapping ± ranges mean the models are statistically tied.
Scroll the table sideways to see every column.
| Rank | Model | Lab | Score | Votes |
|---|---|---|---|---|
| 1 | Claude Fable 5.2 (xHigh) | Anthropic | 1,826 ±19 | 1,445 |
| 2 | Claude Opus 5.5 (Max) | Anthropic | 1,816 ±12 | 4,399 |
| 3 | Claude Sonnet 5.5 (xHigh) | Anthropic | 1,772 ±10 | 6,120 |
| 4 | Gemini 4 Carbon (High)Marked preliminary by LMArena. | 1,768 ±15 | 2,105 | |
| 5 | GPT-6 Astra (Max) | OpenAI | 1,763 ±8 | 8,685 |
| 6 | Qwen4-Max (xHigh)Marked preliminary by LMArena. | Alibaba | 1,750 ±15 | 1,814 |
| 7 | Claude Fable 5.1 (Max) | Anthropic | 1,743 ±8 | 9,990 |
| 8 | GPT-6.1 Sol (Max) | OpenAI | 1,742 ±10 | 5,804 |
SWE-Bench Pro (public): top 6Old: no new results since Jul 2026
Share of real, long-horizon issues from copyleft open-source repositories that a model's patch resolves without breaking existing tests. Higher is better.
Scroll the table sideways to see every column.
| Rank | Model | Lab | Resolved | Added |
|---|---|---|---|---|
| 1 | Muse Spark 1.1* | Meta | 61.5% ±3.1 | 2026-07-09 |
| 2 | GPT-5.4 (xHigh)* | OpenAI | 59.1% ±3.6 | 2026-04-08 |
| 3 | Muse Spark* | Meta | 55.0% ±3.6 | 2026-04-08 |
| 4 | Claude Opus 4.6 (thinking)* | Anthropic | 51.9% ±3.6 | 2026-04-08 |
| 5 | Gemini 3.1 Pro (thinking)* | 46.1% ±3.6 | 2026-04-08 | |
| 6 | Claude Opus 4.5 (2025-11-01) | Anthropic | 45.9% ±3.6 | 2025-12-11 |
SWE-bench Verified, same harness: top 7Old: no new results since Feb 2026
Share of 500 human-validated GitHub issues resolved when every model runs in the same mini-SWE-agent environment. Higher is better.
Scroll the table sideways to see every column.
| Model | Resolved | Reported run cost (500 tasks) | Run date |
|---|---|---|---|
| Claude 4.5 Opus (high) | 76.8% | $376.95 | 2026-02-17 |
| Gemini 3 Flash (high) | 75.8% | $177.98 | 2026-02-17 |
| MiniMax M2.5 (high) | 75.8% | $36.64 | 2026-02-17 |
| Claude 4.6 Opus | 75.6% | $275.76 | 2026-02-17 |
| GLM 5 (high) | 72.8% | $267.19 | 2026-02-17 |
| GPT 5.2 (high) | 72.8% | $236.78 | 2026-02-17 |
| GPT 5.2 Codex | 72.8% | $224.71 | 2026-02-19 |
Agents
Which agent + model pairs finish long, multi-step terminal tasks?
Terminal-Bench 4.1: best run per model, top 12
Share of hard, real terminal tasks an agent harness + model pair completes end to end. Higher is better. Cost is the total the submitter reported for all 330 trials.
Scroll the table sideways to see every column.
| Rank | Agent + model | Effort | Resolved | Total cost (330 trials) | Cost per trial (total ÷ 330) | Date |
|---|---|---|---|---|---|---|
| 1 | Claude Code + Claude Opus 5.5 | max | 65.8% ±3.1 | $4,719 | $14.30 | 2026-09-22 |
| 1 | Claude Code + Claude Fable 5.2 | max | 65.8% ±2.9 | $16,493 | $49.98 | 2026-11-09 |
| 3 | Antigravity CLI + Gemini 4 Carbon | high | 64.6% ±3.2 | $1,358 | $4.12 | 2026-10-28 |
| 4 | Claude Code + Claude Sonnet 5.5 | max | 63.9% ±3.0 | $7,334 | $22.22 | 2026-09-28 |
| 5 | Codex + GPT-6.1 Sol | max | 60.9% ±3.0 | $634 | $1.92 | 2026-09-29 |
| 6 | Codex + GPT-6 Astra | max | 60.3% ±2.8 | $3,267 | $9.90 | 2026-09-03 |
| 8 | Claude Code + Kimi K3.1 Pro | max | 59.7% ±4.2 | $1,087 | $3.29 | 2026-10-24 |
| 12 | Antigravity CLI + Gemini 4 Hydrogen | high | 56.4% ±3.0 | $784 | $2.38 | 2026-11-12 |
| 13 | Antigravity CLI + Gemini 4 Argon | high | 55.5% ±2.9 | $1,109 | $3.36 | 2026-09-30 |
| 17 | Codex + GPT-6.1 Terra | max | 53.3% ±3.1 | $874 | $2.65 | 2026-10-26 |
| 29 | Grok Build + Grok 4.7 | xhigh | 38.8% ±3.6 | $3,683 | $11.16 | 2026-09-21 |
| 30 | Claude Code + DeepSeek V4.1 Flash 1017 | max | 38.5% ±3.8 | $185 | $0.56 | 2026-10-17 |
Search
Which models give the best answers when they search the web?
Search Arena: top 8
People compare two anonymous search-enabled models on the same question and vote for the better answer. Higher is better; overlapping ± ranges mean a statistical tie.
Scroll the table sideways to see every column.
| Rank | Model | Lab | Score | Votes |
|---|---|---|---|---|
| 1 | Qwen3.8-Max | Alibaba | 1,260 ±14 | 1,998 |
| 2 | GPT-5.6 Sol (xHigh) | OpenAI | 1,258 ±6 | 62,364 |
| 3 | Claude Opus 4.6 (Search) | Anthropic | 1,252 ±5 | 166,732 |
| 4 | GPT-5.5 (Search) | OpenAI | 1,246 ±5 | 121,489 |
| 5 | Claude Opus 5 (High) | Anthropic | 1,239 ±7 | 34,000 |
| 6 | Claude Fable 5.1 (High) | Anthropic | 1,237 ±7 | 33,769 |
| 7 | GPT-6 Astra (Max) | OpenAI | 1,231 ±8 | 17,456 |
| 8 | GPT-6.1 Sol (Max) | OpenAI | 1,230 ±8 | 26,556 |
Hallucination
Which models invent the fewest facts?
Vectara hallucination leaderboard: summarisation
How often a model adds facts that are not in a document it was asked to summarise, judged by Vectara's HHEM model. Lower hallucination rate is better.
Scroll the table sideways to see every column.
| Position (of 111) | Model | Hallucination rate | Factual consistency | Answer rate |
|---|---|---|---|---|
| 1 | antgroup/finix_s1_32b | 1.8% | 98.2% | 99.5% |
| 2 | openai/gpt-5.4-nano-2026-03-17 | 3.1% | 96.9% | 100.0% |
| 3 | google/gemini-2.5-flash-lite | 3.3% | 96.7% | 99.5% |
| 6 | deepseek-ai/DeepSeek-V4.1-Flash | 4.2% | 95.8% | 99.9% |
| 30 | openai/gpt-6-sol | 6.5% | 93.5% | 100.0% |
| 45 | deepseek-ai/DeepSeek-V4-Pro | 8.6% | 91.4% | 97.2% |
| 46 | openai/gpt-6-astra | 8.7% | 91.3% | 100.0% |
| 60 | anthropic/claude-opus-5-5 | 9.9% | 90.1% | 99.7% |
| 71 | moonshotai/kimi-k2.6 | 10.8% | 89.2% | 99.7% |
| 85 | google/gemini-4-argon | 12.1% | 87.9% | 99.1% |
| 103 | xai-org/grok-4-1-fast-reasoning | 19.2% | 80.8% | 99.7% |
AA-Omniscience: knowledge and hallucination
6,000 factual questions across 42 topics. Hallucination rate (lower is better) is the share of questions a model didn't get right where it gave a wrong answer instead of declining. Net Score runs from −100 to 100 and rewards correct answers while penalising wrong ones.
Scroll the table sideways to see every column.
| Model (setting) | Hallucination rate | Accuracy | Net Score |
|---|---|---|---|
| Gemini 4 Hydrogen (Xhigh) | 8.5% | 50.5% | 46.3 |
| Gemini 4 Carbon (High) | 11.4% | 58.3% | 53.5 |
| Muse Wonder Full (Max) | 17.6% | 41.0% | 30.6 |
| Grok 4.7 (Xhigh) | 20.0% | 42.7% | 31.2 |
| Qwen4 Max | 28.4% | 43.7% | 27.7 |
| GPT-6.1 Terra (Max) | 34.5% | 47.6% | 29.6 |
| Kimi K3.1 Pro | 37.3% | 48.9% | 29.9 |
| Claude Sonnet 5.5 (Max) | 43.3% | 51.5% | 30.5 |
| GPT-6 Astra (Max) | 49.8% | 61.6% | 42.4 |
| Claude Opus 5.5 (Max) | 50.3% | 65.5% | 48.1 |
| GPT-6.1 Sol (Max) | 51.9% | 61.8% | 42.0 |
| Claude Fable 5.2 (Max) | 59.3% | 79.1% | 66.8 |
Cost per task
What does a task actually cost, and what are the list prices?
Intelligence vs. cost per task
Artificial Analysis Intelligence Index (higher is better) next to the average cost in USD of one task in that index. A cheap model that needs many retries can still cost more per finished job.
Scroll the table sideways to see every column.
| Model (setting) | Intelligence Index | Cost per task |
|---|---|---|
| Claude Fable 5.2 (Max) | 64.6 | $26.54 |
| Gemini 4 Carbon (High) | 60.0 | $5.48 |
| Claude Opus 5.5 (Max) | 59.1 | $10.04 |
| GPT-6 Astra (Max) | 56.9 | $6.24 |
| GPT-6.1 Sol (Max) | 54.6 | $1.27 |
| Claude Sonnet 5.5 (Max) | 54.5 | $10.02 |
| Gemini 4 Hydrogen (Xhigh) | 53.9 | $3.16 |
| GPT-6.1 Terra (Max) | 51.3 | $0.96 |
| Kimi K3.1 Pro | 46.5 | $3.68 |
| Qwen4 Max | 42.9 | $3.27 |
| Claude Haiku 5.5 (Max) | 36.5 | $0.41 |
| Grok 4.7 (Xhigh) | 35.5 | $7.77 |
| DeepSeek V4.1 Flash 1017 (Max) | 32.6 | $0.60 |
Official API list prices
Standard, pay-as-you-go prices per million tokens from each lab's own pricing page. Batch, caching and long-context rates differ; follow the source link for those.
Scroll the table sideways to see every column.
| Model | Lab | Input / 1M tokens | Output / 1M tokens | Source |
|---|---|---|---|---|
| Claude Fable 5.2 | Anthropic | $8.00 | $40.00 | Anthropic |
| Claude Opus 5.5 | Anthropic | $4.00 | $20.00 | Anthropic |
| Claude Sonnet 5.5 | Anthropic | $2.00 | $10.00 | Anthropic |
| Claude Haiku 5.5Prompts up to 100,000 tokens; longer prompts cost more. | Anthropic | $0.10 | $0.50 | Anthropic |
| GPT-6 Astra | OpenAI | $10.00 | $50.00 | OpenAI |
| GPT-6.1 Sol | OpenAI | $2.00 | $10.00 | OpenAI |
| GPT-6.1 Terra | OpenAI | $1.00 | $5.00 | OpenAI |
| GPT-6 Luna | OpenAI | $0.10 | $0.50 | OpenAI |
| Gemini 4 CarbonRises to $4 input / $20 output on 1 December 2026. | $2.00 | $10.00 | ||
| Gemini 4 Hydrogen | $1.50 | $7.50 | ||
| Grok 4.7 | xAI | $2.00 | $6.00 | xAI |
Release tracker
New models (release dates as listed by Artificial Analysis) and new agent results (submission dates on Terminal-Bench 4.1), newest first.
Qwen4 Max · Alibaba · Model
Listed with an Intelligence Index of 42.9. Artificial Analysis
Gemini 4 Hydrogen · Google · Model
Lowest AA-Omniscience hallucination rate in our table (8.5%). Artificial Analysis
Antigravity CLI + Gemini 4 Hydrogen · Google · Agent result
New Terminal-Bench 4.1 entry: 56.4% resolved. Terminal-Bench (Stanford / Laude Institute)
Claude Fable 5.2 · Anthropic · Model
Highest Intelligence Index on the board (64.6 at Max). Artificial Analysis
Claude Code + Claude Fable 5.2 · Anthropic · Agent result
Ties for first on Terminal-Bench 4.1 at 65.8%. Terminal-Bench (Stanford / Laude Institute)
Muse Ember · Meta · Model
New release listed. Artificial Analysis
Tencent Hy4 · Tencent · Model
New release listed. Artificial Analysis
Mistral Large 4 · Mistral · Model
Open-weights release. Artificial Analysis
Gemini 4 Carbon · Google · Model
Intelligence Index 60.0 at High. Artificial Analysis
Antigravity CLI + Gemini 4 Carbon · Google · Agent result
New Terminal-Bench 4.1 entry: 64.6% resolved. Terminal-Bench (Stanford / Laude Institute)
GPT-6.1 Terra · OpenAI · Model
Intelligence Index 51.3 at about $0.96 per task. Artificial Analysis
Kimi K3.1 Pro · Kimi · Model
Open-weights; Intelligence Index 46.5. Artificial Analysis
DeepSeek V4.1 Flash 1017 · DeepSeek · Model
Open-weights; $0.60 per task on the Intelligence Index. Artificial Analysis
Muse Wonder · Meta · Model
New release listed. Artificial Analysis
Claude Haiku 5.5 · Anthropic · Model
Lowest cost per task in our table ($0.41). Artificial Analysis
Gemini 4 Argon · Google · Model
New release listed. Artificial Analysis
GPT-6.1 Sol · OpenAI · Model
Intelligence Index 54.6 at $1.27 per task. Artificial Analysis
Claude Sonnet 5.5 · Anthropic · Model
New release listed. Artificial Analysis
Claude Opus 5.5 · Anthropic · Model
Intelligence Index 59.1 at Max. Artificial Analysis
Sources & method
Figures are copied as published. We round only where noted and never estimate, fill in or combine scores across benchmarks. Ranks are the publisher's own. Tables show a selection of rows, so read the full board at the source before you decide.
Benchmarks measure different things and change over time. A model missing from a table usually means the publisher has not tested it yet, not that it did badly.
Spotted an outdated or wrong number? Tell the Metarix Team via the contact page.
| Source | Publisher | Figures as of | Checked |
|---|---|---|---|
| Code Arena (WebDev) leaderboardElo-style score from blind human votes on front-end and agentic web-dev tasks. | LMArena | ||
| Search Arena leaderboardElo-style score from blind human votes on answers that use live web search. | LMArena | ||
| SWE-Bench Pro (public set)asOf is the newest entry on the board. | Scale AI | ||
| SWE-bench Verified (mini-SWE-agent runs)Figures from the board's published data file (data/leaderboards.json); asOf is the newest Verified entry. | SWE-bench team | ||
| Terminal-Bench 4.1 leaderboardasOf is the newest submission date on the board. Each entry is an agent + model pair run for 330 trials. | Terminal-Bench (Stanford / Laude Institute) | ||
| Hallucination Leaderboard (HHEM) | Vectara | ||
| AA-OmniscienceThe page shows no stamp; asOf is the day it was read. | Artificial Analysis | ||
| LLM leaderboard (Intelligence Index, cost per task, release dates)The page shows no stamp; asOf is the day it was read. | Artificial Analysis | ||
| Claude API pricing | Anthropic | ||
| OpenAI API pricing | OpenAI | ||
| Gemini Developer API pricing | |||
| xAI models and pricing | xAI |