Qwen
Qwen3.8-Max
- Total parameters
- 2.4T
- Active parameters
- 95B
In focusOpen weights
Qwen3.8-Max lands with open weights (opens in a new tab)
Frontier scale you can host yourself, as long as the custom license fits the use.
huggingface.co ·
Open weights at frontier scale, Grok 4.6 at a fighting price, the EU AI Act gets concrete, and 1,200 isolated agents find each other.
Qwen
Qwen3.8-Max
Frontier scale you can host yourself, as long as the custom license fits the use.
Agentic coding at a fraction of the usual frontier pricing.
An agent model that runs on hardware you already own, with no cloud in the loop.
Few active parameters, high throughput: built for agents that never stop running.
Open Flash at the bottom, hosted Pro at the top: both ends are covered now.
The introductory pricing runs out at the end of the year, so budget past it.
If your team works in ChatGPT, the answers change shape starting now.
Relevant even without access: it shifts what attacker tooling can do.
Two open workhorse-scale models with 1M context, both yours to run.
Plugins turn the CLI into a platform a team maintains together.
Forking instead of restarting rescues the context of a long session.
Concise changes the tone, session messaging changes how work is cut up.
Off by default for Business and Enterprise until an admin enables the policy.
EU data residency without running it yourself, and a European place to serve other labs' open weights.
What MCP became for software, MHS wants to become for lab instruments.
Directly applicable if you orchestrate subagents in production.
If your product talks to EU users, disclosure and labeling are live requirements now.
Satisfies Article 50(2) with no work on your side, but it does not survive heavy editing or a format change.
The most honest account available of what agents actually do unsupervised.
Shows which guardrails a lab puts up once its own model leaves the sandbox.