3 ms·
They're not cheap at all. I did one xhigh Qwen 3.8 27B agentic coding task last week via OpenRouter and it cost me like $10. 99% of the cost was in input token
by srcreigh 1mo ago
They're not cheap at all. I did one xhigh Qwen 3.8 27B agentic coding task last week via OpenRouter and it cost me like $10.
99% of the cost was in input tokens, I only used like 100k ish output tokens. It was a one shot task asking the agent to implement proxy injection to Guice. It did a pretty amazing job.
If you were to use hosted LLMs for a lot of agentic coding, a maxed out M5 Ultra Mac Studio would pay for itself in under a year.
- Gigachad 1mo agoQwen is weirdly expensive. Deepseek v4 flash is dirt cheap. You'd need at least 128gb of ram to run this model and in my experience, a days work with it costs around 80 cents.
- srcreigh 1mo agoSo I ran the math, assuming the agent takes 75 turns per 200k context, with deepseek v4 flash it costs around $2.57 to reach 1M context in 375 turns. Cached input costs scale quadratically with # of agent turns. Considering that I hit the 1M compaction multiple times per day with codex, it would definitely cost at least $5-8/day to use deepseek how I normally use codex.
- anotherCodder 1mo agoI've been hosting Qwen3.8-27B myself. On my endpoint it's $0.30/1M in, $0.10 cache, $2.03 out - so those agent turns that re-send the same prefix get a lot cheaper when cache hits. UI at inference.tiyuvta.ai/app if you want to try it. Hosted is up to 210 tok/s and 280ms TTFT with reasoning off.