Sigma Browser

Super Private AI browser with the local AI

Founder: @skinbagwbones

[Sigma Browser @Sigma_Browser]

Hermes agent running in private browser on free local models

Private by design. Local. Open-source

[atomic.chat @atomic_chat_hq]

Nemotron 3 Ultra performed GPT 5.5 level 10× cheaper

We gave three same prompts to build HTML5 canvas with real physics. At first scene we have water in a spinning drum. Galton board - balls through pegs into bins. And a block collision setup with extreme mass differences.

Outputs:
Nemotron 3 Ultra: 11.3k tokens, $0.051
GPT 5.5: 11.0k tokens, $0.57

Nemotron stays right on GPT 5.5's heels, but at 10× cheaper. The gap in quality is far smaller than the gap in price.

[atomic.chat @atomic_chat_hq]

New Google Gemma 4 12B claims near-26B performance - we tested both!

We ran both models locally on one RTX 4090 and gave each the same task: write a self-contained HTML5 canvas animation with real physics in one file without libraries. Three scenes - a Galton board, two blocks colliding off a wall, and a chaotic triple pendulum

Outputs:
Gemma 4 26B-A4B: 15 GB VRAM usage, 6.9k tokens, 138 tok/s
Gemma 4 12B: 9 GB VRAM usage, 8.9k tokens, 80 tok/s

Same Gemma 4 family, but the 26B-A4B won every scene and ran ~1.7x faster - on just 4B active params. The 12B stayed very close though, on almost half the VRAM - which makes it the ideal model for a 16 GB laptop

[Sigma Browser @Sigma_Browser]

@ValerianWaters @DataChaz Honestly haven't benchmarked with 3-4 concurrent agents yet. Real talk though: local agent performance is mostly a RAM story

[Sigma Browser @Sigma_Browser]

@HarryTandy @DataChaz Yep!

[Sigma Browser @Sigma_Browser]

@hanzi_li @DataChaz Especially if you've got a good device

[Sigma Browser @Sigma_Browser]

@Timur_Yessenov @testingcatalog Interesting take. Maybe we should be thinking about a "categorizer" layer for agents on our side too

[Sigma Browser @Sigma_Browser]

@zihua_ma @testingcatalog Exactly! Hermes with its on-the-fly learning makes the in-browser agent genuinely smart

[Sigma Browser @Sigma_Browser]

@sir4K_zen @DanKornas Nothing but privacy!

[Sigma Browser @Sigma_Browser]

@onchaindudes This issue was real in the early versions, but the Nous Research team is actively working on improving the agent - and we ship the latest version of the agent in our browser ;)

[Sigma Browser @Sigma_Browser]

Hermes agent running in private browser on free local models

Private by design. Local. Open-source

[Sigma Browser @Sigma_Browser]

@Tom1217941 All local AI functionality is built into the browser. Also you can choose between Hermes and OpenClaw

[Sigma Browser @Sigma_Browser]

@onchaindudes What's wrong with Hermes?

[Sigma Browser @Sigma_Browser]

@BenSE_ @Teknium The local models component of our previous version is already open source, and we're working on open-sourcing the remaining part

[Sigma Browser @Sigma_Browser]

@0xZroy Using the agent with local models is completely free

[Sigma Browser @Sigma_Browser]

@santoretech Thanks!

[Sigma Browser @Sigma_Browser]

@rohanpaul_ai That's right!

[Sigma Browser @Sigma_Browser]

@schwarzerrttr Surprisingly, Gemma 4B performed quite well for agentic interactions, but if you have more powerful hardware, you can try Qwen 35B

[Sigma Browser @Sigma_Browser]

@NousResearch @Teknium what do you guys think about this?

[Sigma Browser @Sigma_Browser]

Available on MacOS: https://t.co/Uolzb2iKPC

[Sigma Browser @Sigma_Browser]

@Hong60282445 Yes! Remote model support is on the roadmap and coming soon. You'll be able to point your agent at a remote endpoint and run OpenClaw smoothly even on lighter laptops

[Sigma Browser @Sigma_Browser]

Private AI browser with the OpenClaw agent on free local models

Run your agent on Qwen, Gemma, or Nemotron directly in the browser

Open source. Private. Runs on your local device