Changelog: 2026-08-02
We published a new long-form guide comparing self-hosted frontier open-weight models vs API pricing, with clearer breakeven math and refreshed messaging for today’s positioning.
- Released a new deep-dive blog post on when self-hosting frontier open-weight models can outperform API economics.
- Added measured performance and cost comparisons, plus practical guidance on keeping deployments efficiently utilized.
- Corrected and clarified breakeven calculations to avoid misleading throughput assumptions.
- Refreshed positioning, language, and calls-to-action to align with current product messaging.
- Reframed an example “filing screener” workflow as a representative use case rather than a product definition.
- Published a long-form article that walks through the conditions under which self-hosting makes sense, emphasizing the importance of high utilization (“saturation”) for cost-effective operation.
- Included benchmarking-driven discussion of speed and cost tradeoffs, with notes on how different workload patterns can change the economics (for example: output-heavy tasks, fresh-input workloads, and cache-friendly mixes).
- Updated the core breakeven explanation to a more general identity (“breakeven utilization = saturated cost / API price”), and adjusted surrounding narrative so readers can apply it without relying on a single headline throughput number.
- Regenerated the published page output and updated figure captions/footers to match the corrected math and story.
- Tightened the closing section and softened the call-to-action to focus on collaboration (“compare notes”), while aligning the overall framing with the current mission and audience.