whatmodel.sh / Ornith 1.0 35B (35B-A3B)

Ornith 1.0 35B (35B-A3B)

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Ornith 1.0 35B (35B-A3B) is an open-weights language model from DeepReinforce, released June 2026. Its published size is 35B parameters (3B active per token) and a 262k-token context window. Licence: mit.

Catalogue identity notes

DeepReinforce's Ornith 1.0 35B MoE agentic-coding model (ornith-ai/Ornith-1.0-35B @5df2ed3, created 2026-06-21, MIT; the card's links use the former org name deepreinforce-ai, which the Hub redirects to ornith-ai). Reasoning model (opens with <think>). Independent rows: Kwaipilot's in-house column 'Ornith-1.0-35B' in the KAT-Coder-V2.5-Dev card; lab rows: the 'Ornith-1.0-35B-A3B' sibling column of the Ornith 1.5 35B card. ollama 'ornith' (35b tags, updated 3 months ago) is this 1.0 release; 'ornith-1.5' is the later one.

Model information sources

Release date ↗ · Weights licence ↗

Maker: the Ornith 1.0 card lists a '35B-MoE' member ('post-trained on top of Gemma 4 and Qwen 3.5'); the Ornith 1.5 card labels it 'Ornith-1.0-35B-A3B'. ollama reports 34.7B parameters (arch qwen35moe).

67.7±1.1#107 of 719 LLMsfrom 17 benchmark results across 2 sources

$ how it compares

The LLMs ranked either side of Ornith 1.0 35B (35B-A3B), out of 828 with a score, then more from DeepReinforce.

#modelscorewhereRAMreleased
120MiniMax-M2.7 MiniMax68±1api $0.53/M +local~75GB2026-03
121Qwen3.6 35B-A3B Alibaba68±1api $0.31/M +local~23GB2026-04
122Ornith 1.0 35B (35B-A3B) DeepReinforce68±1local~23GB2026-06
123Hy3 preview (295B-A21B) Tencent68±1local~629GB2026-04
124GPT-5 mini OpenAI67±1api $0.69/M2025-08
more from DeepReinforce
83Ornith 1.5 35B-A3B DeepReinforce71±2local~22GB2026-08

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #93 of 300 · 78% of its score · 9 tests

Beats Grok 4.20 (0309 reasoning), Grok 4.3, o3-pro here, though they rank higher overall. Trails Hy3 preview (295B-A21B) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Terminal-Bench 2.1 +136.0% ✎▄······█▄▄█▄ beats 6/7very hard · avg 7.1%▇▇▇▇▇▇▇ 19%at its level · several runs
SWE-bench Verified +155.8% ✎▄▄·····▄▄▄▄█ beats 2/7hard · avg 17%▇▇▇▇▇▇▇ 18%sharp · at its level · several runs
SWE-bench Pro (split unspecified) +134.5% ✎▄···▄···▄▄██ beats 3/7very hard · avg 6.5%▇▇▇▇▇▇▇ 17%sharp · at its level · several runs
SWE-bench Multilingual +151.7% ✎▄·▄·····█▄█▄ beats 4/7hard · avg 18%▇▇▇▇ 10%at its level · several runs
BrowseComp lab63.5% ✎▂·▄▁▂▃▆▄█▂▃▄ beats 28/42hard · avg 23%▇▇ 3.8%at its level · small run
MCP-Atlas (public set) lab64.4% ✎▂·▁▂▁·▁▃█▆▆▂ beats 17/37hard · avg 29%▇ 3.6%at its level · small run
Humanity's Last Exam with tools lab30.1% ✎▂▂▂▁▄▂▃▆█▇▂▆ beats 9/54very hard · avg 13%▇ 2.8%at its level · small run
WideSearch lab63.4% ✎█·▃▆··▂···▂▂ beats 4/10medium · avg 35%▇ 2.5%at its level · small run
1 more test
testresultvs the others in its rundifficultycountswhy
Toolathlon-Verified lab42.4% ✎▂·▂▃▅▃···▂██ beats 2/18hard · avg 27%▇ 1.9%blunt · at its level · small run
coding▁▁▁▃▆██▇▆▄▃▁ #102 of 458 · 50% of its score · 4 tests

Beats Grok 4.3, Muse Glimmer 30B, Qwen3.5 Plus here, though they rank higher overall. Trails Llama 3.1 Nemotron Ultra 253B v1, Solar Pro 4, Nemotron Cascade 2 30B-A3B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
SWE-bench Verified +155.8% ✎▄▄·····▄▄▄▄█ beats 2/7hard · avg 17%▇▇▇▇▇▇▇ 18%sharp · at its level · several runs
SWE-bench Pro (split unspecified) +134.5% ✎▄···▄···▄▄██ beats 3/7very hard · avg 6.5%▇▇▇▇▇▇▇ 17%sharp · at its level · several runs
SWE-bench Multilingual +151.7% ✎▄·▄·····█▄█▄ beats 4/7hard · avg 18%▇▇▇▇ 10%at its level · several runs
SciCode (level unspecified)30.3% ✎▄···▄▄█▄·▄·▄ beats 3/7very hard · avg 13%▇▇ 5.0%at its level
web search█··········█ #2 of 2 · 9.0% of its score · 3 tests
testresultvs the others in its rundifficultycountswhy
BrowseComp lab63.5% ✎▂·▄▁▂▃▆▄█▂▃▄ beats 28/42hard · avg 23%▇▇ 3.8%at its level · small run
Humanity's Last Exam with tools lab30.1% ✎▂▂▂▁▄▂▃▆█▇▂▆ beats 9/54very hard · avg 13%▇ 2.8%at its level · small run
WideSearch lab63.4% ✎█·▃▆··▂···▂▂ beats 4/10medium · avg 35%▇ 2.5%at its level · small run
other tests3 not in a task · 17% of its score
testresultvs the others in its rundifficultycountswhy
GPQA Diamond lab86.2% ✎▁▁▁▁▂▄▅▅███▄ beats 237/310hard · avg 45%▇▇▇ 8.6%sharp · small run
Humanity's Last Exam lab20.8% ✎▁▁▁▃▆▆▇██▆▁▃ beats 49/104very hard · avg 4.8%▇▇ 4.3%small run
NL2Repo lab34.6% ✎▂▂·▂▂▄█▆▆▂▄▆ beats 6/21very hard · avg 11%▇ 3.6%at its level · small run

$ where to run it

llama.cpp · recommended

version

  1. Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
    macOS / Linux
    curl -LsSf https://llama.app/install.sh | sh
    Windows
    irm https://llama.app/install.ps1 | iex

    other ways to install ↗

  2. Run Ornith 1.0 35B (35B-A3B) · needs ~21.2 GB · from huggingface.co/ornith-ai (official)
    chat in the terminal
    llama cli -hf ornith-ai/Ornith-1.0-35B-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf ornith-ai/Ornith-1.0-35B-GGUF:Q4_K_M

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run Ornith 1.0 35B (35B-A3B) · needs ~21.2 GB · from huggingface.co/ornith-ai (official)

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/ornith-ai/Ornith-1.0-35B-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

version

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run Ornith 1.0 35B (35B-A3B) · needs ~21 GB · from ollama.com
    chat in the terminal
    ollama run ornith:35b

MLX (Mac)

version

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run Ornith 1.0 35B (35B-A3B) · needs ~20.4 GB, Mac only · from huggingface.co/lmstudio-community
    chat in the terminal
    mlx_lm.chat --model lmstudio-community/Ornith-1.0-35B-MLX-4bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model lmstudio-community/Ornith-1.0-35B-MLX-4bit

Official weights and documentation: huggingface.co/ornith-ai ↗

$ how the score was worked out · ornith-ai/Ornith-1.5-35B-A3B… 54%, Kwaipilot/KAT-Coder-V2.5-Dev… 46%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
ornith-ai/Ornith-1.5-35B-A3B model card · lab-reported12▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 54%MIT
Kwaipilot/KAT-Coder-V2.5-Dev model card5▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 46%Apache-2.0
prior · pull toward the middle▇ 1%
weighted average= 67.7

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json