utCOMPARE MODELS
Smarter, faster, cheaper.
On one chart.
See how models rate on intelligence, coding and math, how fast they answer, and what they cost. Filter to open-weight models of a size you can run, and highlight the ones you care about.
No scores yet for PingPongRealWorldRL. The chart covers text models rated on LMArena’s leaderboard; image, video and audio models aren’t rated here.
- Open weights
- API only
- License not stated
- Highlighted
Overall rating higher is better · gaps smaller than the ± margin are effectively ties
Overall rating vs speed up and to the right is better and faster
Speeds on an RTX 3060 →Scout’s cloud speed checks haven’t started yet. Once they run, each top-rated model is timed daily on a short fixed prompt. For local speeds, see RTX 3060 tests.
| Model | Intelligence | Coding | Math | Votes | Cloud speed | First token | Price in / out per 1M | Size | Access |
|---|---|---|---|---|---|---|---|---|---|
| Gemini 4 Carbon (High)google | 1,528 | 1,573 | 1,525 | 4.4K | — | — | — | — | API only |
| Gemini 4 Argon (High)google | 1,520 | 1,550 | 1,533 | 11.9K | — | — | $2 / $10 | — | API only |
| Claude Opus 5.5 (High)anthropic | 1,509 | 1,551 | 1,512 | 11.4K | — | — | $4 / $20 | — | API only |
| Muse Wonder (Max)meta | 1,507 | 1,545 | 1,489 | 7.7K | — | — | $4 / $20 | — | API only |
| Claude Fable 5.1 (Max)anthropic | 1,504 | 1,537 | 1,528 | 17.2K | — | — | $10 / $50 | — | API only |
| Claude Fable 5 (High)anthropic | 1,504 | 1,550 | 1,521 | 45.3K | — | — | $10 / $50 | — | API only |
| Claude Opus 4.6 (High)anthropic | 1,503 | 1,548 | 1,517 | 84.5K | — | — | $5 / $25 | — | API only |
| Claude Opus 4.7 (High)anthropic | 1,501 | 1,551 | 1,504 | 65.1K | — | — | $5 / $25 | — | API only |
| Claude Opus 4.6anthropic | 1,496 | 1,546 | 1,503 | 89.3K | — | — | $5 / $25 | — | API only |
| Gemini 3.8 Flash (High)google | 1,496 | 1,530 | 1,524 | 47.8K | — | — | $0.75 / $3.75 | — | API only |
| Muse Spark 1.3 (Max)meta | 1,495 | 1,534 | 1,508 | 22.1K | — | — | $1.25 / $4.25 | — | API only |
| Claude Opus 4.7anthropic | 1,494 | 1,546 | 1,491 | 66.1K | — | — | $5 / $25 | — | API only |
| Muse Spark 1.2 (xHigh)meta | 1,492 | 1,543 | 1,492 | 11.7K | — | — | $1.25 / $4.25 | — | API only |
| Claude Opus 5 (High)anthropic | 1,492 | 1,534 | 1,521 | 73.2K | — | — | $5 / $25 | — | API only |
| Claude Opus 5 (Max)anthropic | 1,491 | 1,528 | 1,515 | 35.5K | — | — | $5 / $25 | — | API only |
| Muse Spark 1.1meta | 1,490 | 1,533 | 1,494 | 46.4K | — | — | $1.25 / $4.25 | — | API only |
| Gemini 3.7 Flash (High)google | 1,490 | 1,520 | 1,503 | 32.3K | — | — | $0.75 / $3.75 | — | API only |
| Muse Sparkmeta | 1,489 | 1,529 | 1,466 | 14.1K | — | — | — | — | API only |
| Kimi K3 (Max)moonshot | 1,488 | 1,540 | 1,499 | 36.8K | — | — | — | 2.8T | Open weightsKimi K3 license |
| Gemini 3.1 Pro Previewgoogle | 1,487 | 1,521 | 1,488 | 132.5K | — | — | $2 / $12 | — | API only |
Ratings: LMArena text leaderboard ↗ (Oct 28, 2026), with style control, from the public dataset under CC BY 4.0. Ratings come from people comparing two anonymous answers and picking the better one; Intelligence is the overall rating, Coding and Math are LMArena’s categories. Cloud speed is measured by Scout as the median of daily timed runs through OpenRouter; local speeds are on the RTX 3060 tests page. Prices are listed API rates from Models.dev. How Scout builds this chart