Mozilla's 91-page report: open-weight AI now only ~4 months behind the frontier
rohanpaul_ai · x · 2026-09-17
Mozilla's new 91-page report finds open-weight AI is only about 4 months behind the frontier — but the economics are inverted.
- By August, 8 of OpenRouter's 10 most-used models by token volume were open-weight, 7 of them Chinese-built; DeepSeek became the first open model to lead the platform in weekly requests.
- Revenue mismatch: open models handled 20% of measured usage but captured only 4% of model-layer revenue, attributed to closed models costing 6x more per call at 90% capability parity. Mozilla notes the revenue data predates its 2026 usage data.
- 79% of surveyed developers use open models, but only 51% of open-model deployments reach production vs 63% for closed models.
- DeepSeek showed major capability jumps may not require fresh pretraining (10 index points, plus 8 from post-training).
- Kimi K3 and Claude Fable 5 showed a 0.72 per-task failure correlation — 'backup' models often fail on the same problems.
Related event: Mozilla Report: Open-Weight Models Now Just ~4 Months Behind Frontier(9 posts)→
More from Models
- Anthropic's server-side rule reignites the case for owning open models over renting AI — alex_verem · 2026-11-02
- Epoch AI: open-weight GLM 5.3 exploited the patch gap to hack 6 Firefox releases at ~$1,000 each — rohanpaul_ai · 2026-11-02
- SemiAnalysis: safety classifiers sandbag scores of Opus 5.5 and GPT-6 Astra — morqon · 2026-11-02
- Liquid AI's 1.7B LFM3 becomes first local model to nail live fixtures lookup on a MacBook — NielsRogge · 2026-11-02
- Claude upgrade-page traffic rebounded to 17.6M in September, near May's 18.6M peak — firstadopter · 2026-11-02
- Open-source GLiNER-PII scrubs 13 identifiers from a discharge note in 550ms on GPU — MaziyarPanahi · 2026-11-02