Qwen 3.8 Max, a new API model, bills $2 per million input tokens and $6 per million output, but its cheapest reliable configuration is not the default. Testing through the Synthorai gateway revealed t…
A developer at Synthorai measured the real-world cost of prompt caching across five agent scenarios, finding that explicit caching with a 1.25x write premium pays off only when the read:write token ra…
A developer measured a 30x cost swing on Google's Gemini 3.6 Flash by adjusting the reasoning_effort dial, with the minimal setting cutting per-call costs 91–97% on most tasks while breaking multi-ste…
Anthropic's Claude Fable 5 model refused mid-tool-call on 11 of 44 coding-agent turns in a Synthorai evaluation, on mundane tasks such as fixing a config default. The refusal truncates tool arguments …
Zhipu's GLM 5.2 open-weight model, available on Synthorai at roughly one-sixth of frontier per-token prices, achieves frontier-level benchmarks but its per-task cost varies by over an order of magnitu…