Is Ampersand state of the art?Pending benchmarks.Announced Oct 20, 2026; no independent results yet.
humans&'s collaborator model, trained to help people and groups reach their individual and collective goals over long, open-ended interactions. Post-trained from GLM 5.3 with reinforcement learning against simulated users from the lab's Persimmon user model. Available to everyone through the lab's Agents API; no open weights.
The details so far
- Lab
- humans&
- Announced
- Oct 20, 2026
- API
- Live · ampersand-v0.1
- Open weights
- Closed
- Parameters
- About 744B total, about 40B active (GLM 5.3 mixture-of-experts)
- Version
- v0.1 (API id ampersand-v0.1)
- Base model
- GLM 5.3: mixture-of-experts, 78 layers, 256 routed experts plus 1 shared, 8 active per token
- Context
- 1,048,576 tokens (GLM 5.3)
- Modalities
- Text in, text out; can post a message, request a tool, or wait without posting
- Training
- Reinforcement learning in long-horizon, multi-party social simulations, with users played by Persimmon (humans&'s user model)
- Access
- Agents API (OpenAI client compatible), open to all developers
- Price (humans&)
- $1.00 per 1M input tokens, $0.20 cached input, $5.20 output
Tags marked unsure are not stated on the lab's own pages.
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
What the lab claims
Self-reported. Shown for context; not used in any verdict.
Lab claim: in multi-month simulations, users collaborating with Ampersand accomplished more than with Fable and other frontier models.
Lab claim (tentative, per the model card): 59.1% on Collaborative Intelligence (teamwork) vs 54.3% for GLM 5.3, and 39.0% cooperative success on CooperBench vs 17.3% for GLM 5.3.
Sources
LABhumans& announcement on Xx.com/humansand/status/2112674550907552137 · Oct 20, 2026LABIntroducing Ampersand (humans& blog)ampersand.humansand.ai/blog/ · Oct 20, 2026LABAmpersand v0.1 model cardampersand.humansand.ai/blog/model-card.html · Oct 20, 2026LABAmpersand API pricingampersand.humansand.ai/docs/PRICING.html · Oct 20, 2026Tracking continues until an independent board lists it.