OpenRouter Tests Quantization Accuracy and Self-Hosting Break-Even
Start here
OpenRouter ran 280 exact-answer tasks against six endpoints serving Qwen3.8 27B on October 20, 2026 and said it did not detect an accuracy difference between fp16, fp8 and fp4; all six endpoints scored between 98.3% and 99.2%. OpenRouter cautions that the test covers one model, 280 tasks and six endpoints on one day, and does not show 4-bit precision is safe for every model or workload. Across 1,455 endpoints, 59.5% report no precision at all, and fp8 is the most common labeled format at 23.6%. Lower precision was cheaper for most models but not all, and provider choice moved price more than precision did: the priciest same-precision endpoint was a median 1.72 times the cheapest. In a separate post, OpenRouter published a break-even analysis using per-provider prices as of October 8, 2026. At those prices, running Qwen3.8 27B on one rented RTX 3090 at $0.50 an hour breaks even at 6 to 12% load, while DeepSeek V4.1 Flash on four H200s never beats the cheapest zero-data-retention endpoint. OpenRouter recommends running a steady base load locally and sending peaks to the API.
Generated from reports · updated 13 h ago
Developments
2 developments
- November 12 08:00 · 1 reportsPublishing a self-hosting vs API break-even analysisOpenRouter Announcements：OpenRouter compares self-hosted LLM and API costs with break-even load calculations
- November 11 08:00 · 1 reportsOpenRouter publishes LLM quantization measurement and routinOpenRouter Announcements：OpenRouter tests what fp8 and fp4 quantization do to LLM quality and price
Report timeline
Follow the reports to see the event from different sides.
- OpenRouter AnnouncementsOpenRouter compares self-hosted LLM and API costs with break-even load calculations
OpenRouter published a break-even analysis showing when self-hosting an open-weight LLM costs less than calling it through an API, using per-provider prices as of October 8, 2026. At those prices, Qwen3.8 27B on one rented RTX 3090 at $0.50 an hour breaks even at 6 to 12% load, while DeepSeek V4.1 Flash on four H200s never beats the cheapest zero-data-retention endpoint. The post also notes that provider.zdr and the EU and US domains cover most privacy needs without self-hosting, and recommends running a steady base load locally while sending peaks to the API.
- OpenRouter AnnouncementsPicksOpenRouter tests what fp8 and fp4 quantization do to LLM quality and price
OpenRouter ran 280 exact-answer tasks against six endpoints of Qwen3.8 27B on October 20, 2026 and did not detect an accuracy difference between fp16, fp8 and fp4, with all six scoring between 98.3% and 99.2%. Across 1,455 endpoints, 59.5% report no precision at all and fp8 is the most common labeled format at 23.6%. Lower precision was cheaper for most models but not all, and provider choice moved price more than precision did, with the priciest same-precision endpoint a median 1.72 times the cheapest.
Heat for this event
Not enough continuous observations yet to draw a trend.