Heavy tier#1 SWE-bench open

Ornith 1.0 35B

DeepReinforce · Ornith

Scores 75.6% on SWE-bench Verified — one of the highest results for any open-weight coding model and above most paid assistants. From the lab DeepReinforce, trained with RL self-improvement for terminal control and tool calling. MIT licensed, 262K context. There is no Ollama library tag yet — pull the GGUF directly with ollama run hf.co/deepreinforce-ai/Ornith-1.0-35B-GGUF.

Params

35B

License

MIT

Context

262K

Min VRAM

20 GB

Min RAM

32 GB

Tier

Heavy

Run locally with Ollama

terminal
ollama run hf.co/deepreinforce-ai/Ornith-1.0-35B-GGUF

Once running, Ollama exposes an OpenAI-compatible endpoint at localhost:11434/v1.

Will it run on my hardware?

Ornith 1.0 35B needs about 20 GB of fast memory at Q4. Here is what in our hardware database clears that bar.

Runs at full speed · 6

NVIDIA RTX 5090 32GB32 GB VRAM~$2,999Amazon
NVIDIA RTX 4090 24GB24 GB VRAM~$2,755Amazon
AMD RX 7900 XTX 24GB24 GB VRAM~$1,339Amazon
Apple Mac mini M4 Pro (24 GB)24 GB unified$1,599Amazon
Apple Mac mini M4 Pro (48 GB)48 GB unified$2,099Amazon
Apple Mac Studio M4 Max (64 GB)64 GB unified$2,899Amazon

Runs, but slower on shared memory · 3

MINISFORUM AI X1 Pro32 GB DDR5 (shared)~$699Amazon
Corsair Vengeance 64 GB DDR5 Kit64 GB DDR5~$1,150Amazon
G.Skill Trident Z5 128 GB DDR5128 GB DDR5~$2,200Amazon

6 other tracked configurations do not have enough memory.

Hardware links are affiliate links. We earn a small commission if you buy through them — at no extra cost to you. Disclosure

Specifications

Parameters35B
Context window262K
LicenseMIT
Min VRAM (Q4)20 GB
Min RAM32 GB
Recommended tierHeavy
Best forcodingdebuggingcode reviewagentic tasks
Reported benchmark75.6% on SWE-bench Verified (as stated by the model developer)

Variants on Ollama

This model has no entry in the Ollama library yet, so it is pulled directly from its HuggingFace GGUF repo.

Ornith 1.0 35B

GGUF pull
ollama run hf.co/deepreinforce-ai/Ornith-1.0-35B-GGUF

Popularity trend (illustrative)

Illustrative adoption trend, not actual download counts. We do not publish a download number for any model because we have no honest source for it.

Similar models