Inkling joins Eval³ leaderboard that scores models on auditing evals for reward hacks
simonguozirui · x · 2026-11-13
Tokenless's Eval³ leaderboard added Thinking Machines' Inkling 2 Small, which reportedly beats other American open-weights. Eval³ has models audit 65 eval/RL environments (up to 200 model calls each) to find reward hacks, with a judge validating findings and pair-wise scoring converted to Elo centered on 1000. Current leaders: GPT-6.1 Sol at 1173 ($335), GPT-6 Astra 1165 ($1,285), Claude Sonnet 5.5 1111 ($454); DeepSeek V4.1 Flash offers standout value at 1069 for just $21, MiniMax M3.1 Flash 986 for $17. Inkling 2 Small ranks 13th at 870 ($38).
More from Models
- Uncensored LFM3 fork drops refusal rate from 61.7% to 0% in a 1.1GB file — alexcovo_eth · 2026-11-13
- Edge0-145B: an open-source 145B-parameter model claiming frontier capabilities on your own computer — BrightCandle · 2026-11-13
- Inkling 2 Small architecture detailed: attention residuals, NVFP4/MXFP8 mixed-precision experts — teortaxesTex · 2026-11-13
- Inkling 2 Small scores 31 on AA index, tops GLM and DeepSeek on finance evals — simonguozirui · 2026-11-13
- Gemini glitch leaks its backend response instructions to the user in plain text — SushiManPlatinum · 2026-11-13
- InferenceBench: agents beat vLLM defaults but lose to simple hyperparameter search — maksym_andr · 2026-11-13