New model performs very well but still burns a ton of tokens
scaling01 · x · 2026-10-18
A brief hands-on note from scaling01: the model in question "performs very well," but still consumes a huge number of tokens — highlighting the ongoing trade-off between capability and inference cost in current frontier models.
More from Models
- Perplexity Decider Flash matches Wald accuracy at nearly 2x speed on M4 Mac — antoine_chaffin · 2026-10-18
- Abliterated DeepSeek V4.1 Flash cuts refusals from 97% to 4% — mrfakename0 · 2026-10-18
- Google team touts progress on Gemini's knowledge work capabilities, citing legal reasoning — aran_nayebi · 2026-10-18
- Gemini tells user it has no web search tool — at Google, the search company — jobergum · 2026-10-18
- DeepInfra cuts DeepSeek V4.1 Flash price 51% as agent demand surges — gharik · 2026-10-18
- SemiAnalysis: vLLM hits 1.3B tokens per $ TCO, 4x better than B300 aggregated serving — woosuk_k · 2026-10-18