Ornith
Shares rank 19 of 39 open-weight language models by recommenders
Local models that people find fast and useful for coding, often ahead of earlier Qwen releases. Critics say coding results are uneven and benchmark claims are unproven.
Local models that people find fast and useful for coding, often ahead of earlier Qwen releases. Critics say coding results are uneven and benchmark claims are unproven.
56 people gave an opinion
- recommend it
- 37
- criticize it
- 19
- approval share
- 66%
Recommended in 5 of 13 months
- November 2025: not recommended
- December 2025: not recommended
- January 2026: not recommended
- February 2026: not recommended
- March 2026: not recommended
- April 2026: not recommended
- May 2026: not recommended
- June 2026: recommended
- July 2026: recommended
- August 2026: recommended
- September 2026: recommended
- October 2026: recommended
- November 2026: not recommended (month in progress)
Why it doesn’t carry the badge
- 37 people recommend itNeeds 10 or more, met.
- 1.9 recommenders for every criticNeeds 2 to 1 or better, not met.
- Recommended in 5 of 13 monthsNeeds 4 or more, met.
The same three rules apply to every product, and the badge can’t be bought. How the badge works
What people say about it
The reasons people give, with the number of comments that give each one. Open a reason to read what they said.
Why people recommend it
Each mark is one comment
Good at coding tasks
“Based on my limited usage, it is good, gives creative solutions to coding problems.”
Comment from 29 Jun 2026 in “Ornith-1.0: self-improving open-source models for agentic coding” “which has performed slightly better and faster at Python, db and some backend tasks”
Comment from 3 Aug 2026 in “Qwen3.8-Max: A New Bar for Coding and Cowork” “it's usable for real coding and sys admin work, and works great in opencode.”
Comment from 15 Oct 2026 in “OpenCode 2.0” - Comment from 3 Aug 2026 in “Qwen3.8-Max: A New Bar for Coding and Cowork”
Beats earlier Qwen models
“it's definitely better than Qwen3.6 35b-a3b which was my go-to before”
Comment from 28 Aug 2026 in “Run Qwen3.8 27B locally: real numbers from my Mac Studio” “I found Ornith1.5 to be a better MoE model for me, also very fast”
Comment from 18 Sep 2026 in “Bonsai 2 27B: Near-Lossless Compression in a 9x Smaller Footprint” “It's been a pretty decent step up for me compared to Qwen3.6”
Comment from 25 Aug 2026 in “Qwen 3.8-Flash-Next releasing tomorrow (125B a6B)” Fast on local hardware
“playing with ornith-35B with a draft vibethinker-3b as a draft gave me some good speed/results”
Comment from 29 Jun 2026 in “Qwen 3.6 27B is the sweet spot for local development” “I found Ornith1.5 to be a better MoE model for me, also very fast”
Comment from 18 Sep 2026 in “Bonsai 2 27B: Near-Lossless Compression in a 9x Smaller Footprint” “much faster on my hardware but better responses”
Comment from 4 Oct 2026 in “Run Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/s” - Comment from 30 Jun 2026 in “Ornith-1.0: self-improving open-source models for agentic coding”
Useful for everyday tasks
“it's pretty nice to just let it work away at simpler prompts overnight or during lunch break”
Comment from 29 Aug 2026 in “Run Qwen3.8 27B locally: real numbers from my Mac Studio” “Ornith 1.0 had problems with world knowledge”
Comment from 29 Aug 2026 in “Run Qwen3.8 27B locally: real numbers from my Mac Studio” “so far its really impressive for my needs”
Comment from 20 Aug 2026 in “Ornith-1.5: From Self-Scaffolding to Self-Improvement” - Comment from 30 Jun 2026 in “Ornith-1.0: self-improving open-source models for agentic coding”
What people warn about
Each mark is one comment
Struggles on some coding tasks
“it reached its limits totally botching a longer (simple) kernel implementation job”
Comment from 30 Jun 2026 in “Ornith-1.0: self-improving open-source models for agentic coding” “It got rather tangled up when I tried it with one of my coding tests”
Comment from 29 Jun 2026 in “Qwen 3.6 27B is the sweet spot for local development” “the Qwen-based 35b MoE Ornith version performs worse than Qwen 3.6 and Qwen AgentWorld on my benchmarks”
Comment from 29 Jun 2026 in “Qwen 3.6 27B is the sweet spot for local development” - Comment from 28 Aug 2026 in “Run Qwen3.8 27B locally: real numbers from my Mac Studio”
- Comment from 29 Jun 2026 in “Ornith-1.0: self-improving open-source models for agentic coding”
Benchmark claims questioned
“they just put graphs with different color prioritizing themselves”
Comment from 29 Jun 2026 in “Ornith-1.0: self-improving open-source models for agentic coding” “These are simply benchmaxxed versions of either Qwen or Gemma 4.”
Comment from 29 Jun 2026 in “Ornith-1.0: self-improving open-source models for agentic coding” “there are no benchmarks or weights for it anywhere”
Comment from 29 Jun 2026 in “Ornith-1.0: self-improving open-source models for agentic coding” - Comment from 6 Jul 2026 in “Right to Local Intelligence”
Loops and hallucinates in tool calls
“Long session tool calls sucks and hallucinate a lot with that too.”
Comment from 29 Jun 2026 in “Ornith-1.0: self-improving open-source models for agentic coding” “seems to get in the loop often specially with tool calls.”
Comment from 2 Sep 2026 in “My local model setup on an M4 Pro Mac Mini” Worse than Qwen 3.6 in tests
“the Qwen-based 35b MoE Ornith version performs worse than Qwen 3.6 and Qwen AgentWorld on my benchmarks”
Comment from 29 Jun 2026 in “Qwen 3.6 27B is the sweet spot for local development” “it performed worse than Qwen3.6-27b in my basic test”
Comment from 29 Jun 2026 in “Ornith-1.0: self-improving open-source models for agentic coding”
Other open-weight language models people recommend
The four most recommended other open-weight language models, with the same counts for each, so you can compare like for like. The tick on each bar marks the 2 to 1 line the badge needs.
| Product | Recommend | Criticize | Approval | Months | Owned | Badge |
|---|---|---|---|---|---|---|
| Qwen 3.8 27B | 1,173 | 426 | 73% | 13 | 2 mo32 owners | Approved |
| Z.ai GLM | 1,043 | 384 | 73% | 13 | 2 mo52 owners | Approved |
| Kimi | 931 | 359 | 72% | 13 | 1 mo26 owners | Approved |
| Gemma 4 | 713 | 249 | 74% | 13 | 1 mo14 owners | Approved |
| OrnithThis page | 37 | 19 | 66% | 5 | — | — |
What people said
40 of the 67 comments that mention it, newest first. Each links to the original comment.
“it's usable for real coding and sys admin work, and works great in opencode.”
“I run these two Qwen model variations successfully.”
“much faster on my hardware but better responses”
“an overall stronger model on the same architecture, so a drop-in replacement”
“I found Ornith1.5 to be a better MoE model for me, also very fast”
“Depending on which benchmarks you believe, these models (and the Ornith 1.5 finetune of Qwen 35B-A3B) are competitive at about Opus 4.5 to 4.7 level.”
“seems to get in the loop often specially with tool calls.”
“it's pretty nice to just let it work away at simpler prompts overnight or during lunch break”
“It's now my go-to model for quick inquiries.”
“Ornith 1.0 had problems with world knowledge”
“I didn’t get very good results coding in Python, Racket, and TypeScript.”
“it's definitely better than Qwen3.6 35b-a3b which was my go-to before”
“It's been a pretty decent step up for me compared to Qwen3.6”
“so far its really impressive for my needs”
“it's been very promising using 32k context”
“Quick tests suggest it's pretty good at reasoning and tool use”
“it has been on par with Qwen3.8 27B at a much higher speed and at a higher quant”
“With the right assistance from Qwen3.5/Ornith I was able to achieve some amazing results”
“which has performed slightly better and faster at Python, db and some backend tasks”
“excels in coding, at least for my coding needs”
“the subtlety and nuance it showed in its response was very surprising to me”
“I've tested extensively with local models such as qwen3.6:27b and ornith:35b, and the results are satisfactory.”
“it was still outperforming a previous generalist model I had running locally”
“The models didn't live up to the claims they made.”
“it reached its limits totally botching a longer (simple) kernel implementation job”
“This one actually felt useful.”
“the model is way faster than Qwen3.6 35B”
“the Qwen-based 35b MoE Ornith version performs worse than Qwen 3.6 and Qwen AgentWorld on my benchmarks”
“it performed worse than Qwen3.6-27b in my basic test”
“they just put graphs with different color prioritizing themselves”
“Long session tool calls sucks and hallucinate a lot with that too.”
“there are no benchmarks or weights for it anywhere”
“Based on my limited usage, it is good, gives creative solutions to coding problems.”
“These are simply benchmaxxed versions of either Qwen or Gemma 4.”
“It is also available with Ollama now and I am equally impressed too.”
“playing with ornith-35B with a draft vibethinker-3b as a draft gave me some good speed/results”
“It got rather tangled up when I tried it with one of my coding tests”
“In my brief tests, Ornith 35B performed quite well.”
“my own silly tests either suggest otherwise or suggest that I have to talk to it a bit differently”
“It performs poorly with only read/grep/ls tools, but in a follow-up test with a full shell and Python, it doubled its findings”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.