Homan 2.96: Our Most Advanced Model Ever
Homan 2.96 is the culmination of everything Bloom Labs has learned about large-scale reasoning systems. It is, by a wide margin, the most capable model we have ever built -- engineered for enterprises and power users who refuse to settle for "good enough." Internal development and testing are complete, but Homan 2.96 has not been released publicly. Given how capable it is, we are holding it back deliberately: our safety team is running a full risk review, and we will not ship it until we are confident it cannot cause harm.
Read the Benchmark Report » Compare All ModelsBuilt for the Hardest Problems You Have
Homan 2.96 was trained from the ground up to push past every ceiling our earlier models ran into. This isn't an incremental update -- it's a categorical leap forward, and it shows in everything the model touches:
- State-of-the-art reasoning. Homan 2.96 chains multi-step logic, weighs tradeoffs, and arrives at defensible conclusions on problems that would stump earlier-generation systems entirely.
- Elite coding ability. From scaffolding entire applications to refactoring legacy codebases, Homan 2.96 writes, debugs, and explains code with a fluency that rivals senior engineers.
- Multimodal understanding. Text, images, diagrams, and screenshots are all first-class citizens -- Homan 2.96 reasons across modalities natively, not as a bolted-on afterthought.
- Massive long-context comprehension. Feed it entire codebases, legal contracts, or research libraries and Homan 2.96 holds the full picture in mind, tracking detail across hundreds of pages without losing the thread.
Simply put: Homan 2.96 is the model we built for the future we saw coming. It is not yet available outside Bloom Labs -- leadership is genuinely concerned that a model this capable could pose real risks if released before it's ready, so it stays behind closed doors until our safety team signs off.
Benchmark Highlights
We are beyond proud of these results. For the full story -- including a genuine breakthrough moment for the team -- see the complete benchmark report linked below.
That multiplication score is our proudest number on this whole page — the first time any Bloom Labs model has reliably done real arithmetic, and by far our highest mark in this suite. The team is thrilled.
Spec Sheet
| Spec | Homan 2.96 |
|---|---|
| Parameters (est.) | 8B |
| Context Window | 256K tokens |
| Modality | Text, Image, Diagram, Code |
| Availability | Coming Soon |
| Starting Price | $0.049 / 1K tokens |
From the Research Blog
Curious how Homan 2.96 stacks up under the hood? Our research team put it through its paces -- read the full writeup, including the historic milestone that has our whole team celebrating.
- Homan 2.96 Benchmark Report: Reasoning, Code, and the Curious Case of Multiplication — Sept 15, 2026
- Introducing Homan 2.96: Our Most Capable Model Yet — Sept 12, 2026