Speculation: OpenAI's efficient models free compute for bigger ones; Opus 5.5 uses 4x output tokens
haider1 · x · 2026-09-25
The author speculates OpenAI's efficient gpt-6 sol/luna models were chosen to free data center capacity for a larger unreleased model (bigger than astra). Notably, Opus 5.5 uses 4x more output tokens — if API pricing roughly reflects model size, that could mean up to 8x more compute at those benchmark scores. Unverified speculation.
More from Infra
- Goodwin raises $720M Series C to build the fastest chips for AI inference scaling — igarciacamargo · 2026-11-12
- PufferLib 5.0 trains a Rubik's Cube solver at 8.8M SPS on a single 12GB GPU — haydendevs · 2026-11-11
- e-means quantization shrinks Qwen3.5-4B to 1.23GB while keeping 91.9% performance — kastnerkyle · 2026-11-11
- Unsloth ships Qwen3.8-27B Dynamic V3.1 GGUFs with 5-10% accuracy gains — danielhanchen · 2026-11-11
- STE regularization compresses ColBERT 5x with 99.4% retrieval retention — antoine_chaffin · 2026-11-11
- Home-built GPU rig claimed to run DeepSeek at 700 tokens/sec — philipvollet · 2026-11-11