New TB-fn Benchmark Reveals Significant Drops in Terminal Model Rankings
abeirami · x · 2026-08-27
Questions about the actual capabilities of top models on the Terminal Bench 2.1 leaderboard led to the creation of the TB-fn benchmark. TB-fn reworks the same 89 tasks by adding steps, tightening requirements, or necessitating more complex algorithms, while removing shortcut solutions. Results show models clustered on TB-2.1 spread across more tiers on TB-fn, with GPT-5.6 Sol and Opus 5 remaining in the lead, while others drop significantly. Models work 38% harder on average in TB-fn.
More from Models
- Unverified: internal snapshot labeled gpt-6-astra-api-priv-ev3 spotted on model tracker — lyraxana · 2026-10-16
- NVIDIA releases Nemotron 3.5 Bolt 74B MoE that fits in 24GB VRAM at 2-bit — NerdyRodent · 2026-10-16
- Muse Code 1.4.8 cuts token usage; all user quotas reset immediately — alexandr_wang · 2026-10-16
- Alibaba and Baidu to power key parts of Apple's AI services in China — nordicinst · 2026-10-16
- Cornell and Amazon revisit the 16x pretraining compute gap: diffusion LMs need only 3-8x — LucaAmb · 2026-10-16
- Image Decision Benchmark JevImageBench hits v0.3.0 with reshuffled Top 5 — airesearch12 · 2026-10-16