Ornith 1.5 122B
GoodThe maker claims 84.5% on SWE-bench Verified. We could not confirm what this model really is. Test it before you trust it.
Same as Maker claims about as strong as Beam, unconfirmed
Made by ornith-ai
About
Ornith 1.5 122B is a 122B coding model released around 20 Oct 2026 under a license the maker calls MIT. The maker publishes it on Hugging Face and Ollama. Every coding score it shows comes from the maker's own chart, and no outside test has confirmed them.
Best for
- Writing code
- Coding agents
Scores on record
- SWE-bench Verified (vendor)84.5%
- TerminalBench 2.1 (vendor)81.2%
Watch out
- People have accused the maker of renaming an Alibaba model. They point to the same architecture and tokenizer as Qwen3.5-122B-A10B. The maker has not confirmed or denied it.
- Every score is the maker's own and no outside test has confirmed it. Treat the numbers with care.
- The company that hosts it does not say what it keeps of your data, so do not send it secrets.
All routes
Nous Portal
free roster LiveFree on Nous Portal. Listed on the free roster tracked by llmroster.dev.
Last checked 24 Oct 2026 at 18:30 IST
Use it free in 1 minute
- Open Nous Portal.
- Sign in, or ask us and we will walk you through it.
- Pick the model and send a test message.
- If it asks for a key, add yours in place of YOUR_KEY.
Command Code
free listing EndedFree until 22 Oct, about 11:30 IST. Now $0.16 in and $0.64 out per 1M. People think this is a renamed Qwen3.5, but nobody has checked. Do not send it secrets.
Last checked 24 Oct 2026 at 18:30 IST
History
No probe checks yet. No checks have run on this model, so there is nothing to show here. For now this model is checked by hand.