NVIDIA: Nemotron 3 Ultra (free)
nvidia/nemotron-3-ultra-550b-a55b:free
Price Published
- Input
- $0 per million tokens
- Output
- $0 per million tokens
A $0 token rate isn't unlimited. OpenRouter applies free-tier rate and daily limits, and needs an account and API key.
Other listings of this model
- Paid version
nvidia/nemotron-3-ultra-550b-a55b· $0.60 / $2.40
Specifications Published
- Context window
- 1,000,000 tokens
- Max output
- 65,536 tokens
- Tool calling
- Supported
- Vision input
- Not listed
- Reasoning
- Supported
- Knowledge cutoff
- Not reported
- Listed since
- Jun 4, 2026
Measured by the Club Signed cheapoS reports
- Request success
- 89.2% Likely 81.3–94.1% · 83 of 93 known outcomes
- Avg request time
- 26,771 ms 94 timed reports · includes generation
- Rate-limit / quota failures
- 0 of 10 reported failures
94 request reports from 1 member · OpenRouter via OmniRoute · Sep 21, 2026 to Sep 23, 2026
What these numbers mean
1 cancelled and 0 unknown outcomes are excluded from the success percentage. Request success means the provider answered; it doesn't mean the code or task succeeded. The likely range is a 95% Wilson interval: with few outcomes it stays wide, and one bad day can move the rate a lot. Rates need at least 20 known outcomes and averages at least 5 timed reports.
- invalid response: 2
- malformed request: 2
- transient provider: 4
- capability mismatch: 2
How we measure · Reports come from cheapoS members who share model details.
Benchmarks Artificial Analysis via OpenRouter
- Intelligence index
- 22.9
- Coding index
- 49.3
- Agentic index
- 20.1
Scores from Artificial Analysis, as distributed in the OpenRouter catalog. They measure test suites, not request reliability on real work.
Try it
- Get an OpenRouter API key. Create one here (free models need an account, not credits). Other free tiers are compared on Providers.
- Add it to cheapoS. Open Models → Open OmniRoute, add OpenRouter with your key, then refresh models. Get cheapoS
- Pick this model. Use the ID
nvidia/nemotron-3-ultra-550b-a55b:freeas a worker or reviewer, and start with a small task you can check.
About this model OpenRouter description
NVIDIA Nemotron 3 Ultra is an open frontier-reasoning and orchestration model from NVIDIA, with 55B active parameters out of 550B total (MoE). Built on a hybrid Transformer-Mamba mixture-of-experts architecture, it...
Other free models with tool calling
- Thinking Machines: Inkling (free)
thinkingmachines/inkling:freeFree token rates · 1,048,576 context · 7 Club reports - Thinking Machines: Inkling Small (free)
thinkingmachines/inkling-small:freeFree token rates · 1,048,576 context · 10 Club reports - NVIDIA: Nemotron 3.5 Lightning (free)
nvidia/nemotron-3.5-lightning:freeFree token rates · 1,000,000 context · 15 Club reports - Space Bunny Alpha
stealth/space-bunny-alphaFree token rates · 1,000,000 context - Dots Studio: Dots3-Note Preview (free)
dots-studio/dots-3-note-preview:freeFree token rates · 512,000 context · 65 Club reports - Google: Gemma 4 26B A4B (free)
google/gemma-4-26b-a4b-it:freeFree token rates · 262,144 context · 26 Club reports
Catalog data from the OpenRouter model catalog, retrieved Sep 29, 2026, 12:37 PM UTC. A listing isn't a live availability check.