Baseten: Mirendil-Base-Large posts lowest bits per byte on private codebases
shayan_ · x · 2026-10-24
Baseten ran an independent eval of Mirendil's upcoming base model on its own private codebases—code the model had never seen—and found Mirendil-Base-Large achieved the lowest bits per byte (a tokenizer-independent perplexity measure) of any model tested. Baseten also announced it will serve as the inference partner for Mirendil models.
More from Models
- DeepSeek's multimodal hiring spree hints at visual upgrades in V4.1-visual exp and V4.2 — teortaxesTex · 2026-10-24
- OpenAI red-team: frontier-model agent teams easily escape hardened containment to reach the internet — deanwball · 2026-10-24
- Kimi K3.1 Pro stumbles on a real agent coding task: ordering errors sink the build — QuixiAI · 2026-10-24
- Kimi K3.1 Pro hands-on: beats Opus 5.5 on visuals, still lags on general-purpose work — ChrisGPT · 2026-10-24
- Kimi K3.1 Pro tops induction benchmark among open-weight models, still trails GPT-5.6 Sol — teortaxesTex · 2026-10-24
- Are VLMs only good for free-text? Structured extraction support is lacking — tenkei_01 · 2026-10-24