Anyone tested this model? Its an MoE made for coding and comes in several sizes… [image] Ornith-1.0: Self-Scaffolding LLMs for Agentic Coding Introducing Ornith-1.0, a self-improving family of open-source models specially for agentic coding tasks. [i…

It seems someone is uploading for spark 4 node CosmicRaisins/Ornith-1.0-397B-W4A16 · Hugging Face . probably 2 nodes would work int4 autoround version comes.

Should be similar recipe to this: New release: Nex-N2-Pro, a 397B parameter MoE model based on Qwen - #23 by karol.spark

[image] Ornith-1.0-397B Released – Has Anyone Tested It Yet or Found the Best Inference Settings? DGX Spark / GB10 It looks like deepreinforce-ai/Ornith-1.0-397B has been released. From what I can tell, it also appears to be a fine-tuned version of the Qwen 3.…

I’ve used its 35B model for driver development, and it wasn’t very good, similar to qwen 3.6 35B-A3B. It can’t even handle skills and tool calls reliably. In comparison, it’s much worse than qwen 3.6 27B. Maybe it’s better suited for higher-level application development or similar tasks. I’m not s…

They called it coding, but it looks like it’s more trained on tool use, both 9b and 35b are pretty good with tools and instruction following. 27b actives will of course beat 3b and 9b actives) but both 9b and 35b are very fast. Best use probably as an agent in a harness like hermes doing sysadmin wo…

interesting, i am using it with hermes @ bf16 (only quant available at the time), and i have found that its skill use and tool calling were a little bit better than the qwen/qwen-35b-fp8… totally subjective assessment.

Tried the 35b variant (FP8) with a complex hermes skill (fundamental analysis of stocks). Compared to qwen3.5-122b – which runs as fast as Ornith – the results were rather poor. For me, it was a waste of time.

Ornith 1.0 Anyone?

Accelerated Computing DGX Spark / GB10 User Forum DGX Spark / GB10 Projects

gpieceoffice June 28, 2026, 6:09am 4

I plan to post the inference commands and benchmark results for the 397B model here.

I’ve already completed a pure RTN-based W4A16G128 quantization using Intel AutoRound and am currently testing it. If everything goes well, I also plan to upload the quantized model to Hugging Face.

If you’re interested in the 397B model, feel free to follow this thread for updates!

Topic		Replies	Views
Ornith-1.0-397B Released – Has Anyone Tested It Yet or Found the Best Inference Settings? DGX Spark / GB10 ai-model-training	22	6941	July 3, 2026
Ornith-1.0-35B-int4-AutoRound For GB10 DGX Spark / GB10	6	1491	June 29, 2026
Single-Spark setups — which models do you actually run for coding, and how? (sharing mine + a test prompt) DGX Spark / GB10 spark , llama , agentic-ai , deepseek , dgx	37	3901	July 9, 2026
Are there plans to add the newly released Ornith-1.0 Models nim , agentic-ai	2	946	June 28, 2026
Best models/configurations for agentic coding with DGX (Nvidia/Asus/Dell/Lenovo/MSI) DGX Spark / GB10 agentic-ai	2	2083	April 17, 2026
Introducing Spark Auto Round /w OpenCode Instruct dataset DGX Spark / GB10 cuda , spark , agentic-ai	110	4079	July 17, 2026
New model (one of the best i ever used, in my humble opinion): Qwen Agent World for local programming DGX Spark / GB10 Projects agentic-ai , dgx	0	538	June 29, 2026
What model can I run on 2 DGX Spark? DGX Spark / GB10 llama , agentic-ai	5	535	July 1, 2026
A Promising (?) Model Trained on Top of Qwen 3.5 397B DGX Spark / GB10	3	880	June 10, 2026
Local coding agent with Asus DX10 DGX Spark / GB10	1	356	June 12, 2026

Ornith 1.0 Anyone?

Related topics