Ornith 1.5 Release - Toolathlon-Verified: leaderboard
Ornith-1.5-122B official release card (2026): source-reported same-table comparisons, missing values stay absent. TerminalBench 2.1 uses Claude Code 2.1.126, temperature/top-p 1 and 131072 output tokens, five runs. SWE variants use OpenHands, temperature 1/top-p 0.95, 256K context and anti-hacking safeguards. NL2Repo uses 400K context and 48K output; HLE uses Claude Opus 4.6 judge; MCP-Atlas uses thinking mode, public 500-task subset and 10-minute limit. Toolathlon uses official Verified service at 128K; ClawEval temperature 0.6 and 256K context. No comparison is fused with a differently configured source
Metric: Score (%). Source: huggingface.co. Saturation forecast: Around January 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen3.8-Flash-Next | 71.4 | #72 |
| 2 | DeepSeek V4 Flash (0731) | 70.3 | #78 |
| 3 | Qwen 3.8 27B | 67.1 | #104 |
| 4 | Nemotron 3.5 Super VL | 60.1 | #167 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=ornith-1-5-release-toolathlon-verified · How It Works · Data refreshed daily, snapshot 2026-10-21.