Benchmarks · Cost
T3 Code output tokens per turn
How much a model writes per turn in real coding work: output tokens, thinking included, per T3 Code turn over the last 30 days. Lower is terser. From T3 Code's telemetry via Slopalytics.
13 models · 13 results · updated 2026-11-12 · slopalytics.com ↗
Leaderboard
GPT-6.1 Terra max
All results
13 rows| # | Model | Effort | Score |
|---|---|---|---|
| 1 | GPT-6.1 Terra | max | 5424.8 |
| 2 | GPT-6 Astra | max | 8575.3 |
| 3 | GPT-5.6 Luna | max | 8611.0 |
| 4 | GPT-5.6 Sol | max | 9033.8 |
| 5 | Opus 5.5 | max | 9141.0 |
| 6 | Fable 5.1 | max | 9734.1 |
| 7 | Opus 5 | max | 10615.3 |
| 8 | Sonnet 5.5 | max | 10629.3 |
| 9 | GPT-6 Sol | max | 11280.0 |
| 10 | GPT-6.1 Sol | max | 12292.3 |
| 11 | Fable 5.2 | max | 12906.9 |
| 12 | Haiku 5.5 | max | 13710.9 |
| 13 | GPT-6 Luna | max | 16224.6 |