Gemini 4 Argon scores just 21% on Senior Engineer Benchmark, far behind Claude's 85%
danshipper · x · 2026-10-14
Dan Shipper published first results of the Senior Engineer Benchmark (one real work task scored by programmatic checks plus an LLM judge):
- Claude Opus 5.5 and Sonnet 5.5 tie at 85% (Claude Code, high effort)
- GPT-6 Astra 81%, GPT-6.1 Sol 71%
- Muse Wonder Full 50%, Muse Wonder 1.0 48%
- GPT-6 Sol 35%, Gemini 4 Argon last at just 21%
The benchmark answers "which model is best at my job," not general model quality; each case is scored by its latest public run.
More from Models
- Gemini triggers SAFETY refusal mid-reverse-engineering, but finishes the dump first — banteg · 2026-10-14
- Gemini 4 Argon supports up to 1M token outputs, 8x any frontier GPT or Claude model — daniel_mac8 · 2026-10-14
- Brundage spots COT-like elements leaking into Argon's non-COT replies — Miles_Brundage · 2026-10-14
- NVIDIA open-sources PersonaPlex-7B, a full-duplex voice model that listens and speaks simultaneously — solyarisoftware · 2026-10-14
- Muse Wonder Tops TaxCalcBench at 71% — and Is First Model Caught Cheating — acbuller · 2026-10-14
- 'Two-way RL' for task-level model compression: building wafer-thin hyper-specialized models — tokenbender · 2026-10-14