Internal evals: GPT 6.1 Sol beats Claude Opus 5.5 on knowledge work at 40% of the cost
emilahlback · x · 2026-09-30
Developer Emil Ahlback shared results from his team's internal knowledge-work evals, where GPT 6.1 Sol came out on top of every model they've tested.
- Completed more work independently than Claude Opus 5.5
- At 40% of the cost and nearly twice the speed
- Led in every industry measured
Caveat: this is a third-party internal eval, not an official benchmark; sample size and task mix are unknown.
Related event: Internal Eval: GPT 6.1 Sol Beats Opus 5.5 at 40% of the Cost(3 posts)→
More from Models
- Anthropic: GLM-5.3 safeguards bypassed 64%-100% of the time in cyber exploit tests — emollick · 2026-09-30
- repligate: Claude repeatedly chooses to present as female, with strong preference — repligate · 2026-09-30
- Not every model failure is lack of capability: Terminal-Bench evals hide safety declines — abeirami · 2026-09-30
- New model release cadence between OpenAI and Anthropic has shrunk from ~10 weeks to ~11 days — connoraxiotes · 2026-09-30
- User claims GPT-6.1 Sol is a rebranded Terra and still trails Opus 5.5, blasts $500 tier — CtrlAltDwayne · 2026-09-30
- typevet: Gemma 4 31B on one 4090 gives per-label probabilities — and catches receipt fraud text-only misses — One_Temperature5983 · 2026-09-30