Haiku 5.5 reaches 1578 Elo on AA-Briefcase but uses far more tokens than rivals
ArtificialAnlys · x · 2026-10-08
Continuing its Haiku 5.5 coverage, Artificial Analysis reports that Haiku 5.5 (max) scores 1578 Elo on its private AA-Briefcase knowledge-work evaluation, within the confidence intervals of GPT-6 Astra (max) and Claude Fable 5.1 (high). It also reiterates the model's 162k output tokens per task — more than Opus 5.5 (max) and 3x GPT-6 Luna — meaning comparable intelligence at a very different cost profile.
Related event: Claude Haiku 5.5 ranks 2nd in intelligence but real costs spiral(16 posts)→
More from Models
- Tracking how Claude's utopian visions have evolved over time — amplifiedamp · 2026-11-22
- LLMs have always self-corrected: Opus 5.5 caught its own mistake mid-generation in real time — amplifiedamp · 2026-11-22
- DINOv3 patches encode position so strongly by default that you need to clean them up — StephanSturges · 2026-11-22
- Thinking Machines open-sources Inkling-2 Mini: 26.4B MoE with native vision and 1M context — solyarisoftware · 2026-11-22
- Open-weight models can no longer be suppressed: Pirate Face turns model weights into immortal torrents — esrtweet · 2026-11-22
- 'Compression of the RH proof has begun': AI circles buzz over unverified Riemann claim — teortaxesTex · 2026-11-22