Ant Group’s inclusionAI launches Ling 3.1 Flash, a 560B-parameter MoE, as a free trial with open weights promised
Partly confirmed
The takeaway
Around Sept 29–30, 2026 Ant Group’s AI unit inclusionAI released Ling 3.1 Flash, a hybrid-reasoning mixture-of-experts model with 560B total and 25B active parameters, aimed at coding, agents and tool use.
Status
- Claim
Partly confirmed
- Our reporting
- Medium confidence
- Importance
- 3 of 5
- Last verified
- 14 October 2026
Your AI and this story
- GPT-6.2 Astra121 days after its cutoff
- Claude Opus 5.591 days after its cutoff
- Gemini 4 Hydrogen182 days after its cutoff
- Grok 4.7121 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 91 days before it.
Key facts
- 560B total / 25B active parameters, hybrid reasoning MoE (reasoning on by default on OpenRouter), text in/text out
- Trial: 262,144-token context, 32,768 max output; free on Vercel AI Gateway through Oct 13, 2026 (release date listed 09/29/2026) and free on OpenRouter (listed Oct 2, 2026) as inclusionai/ling-3.1-flash
- inclusionAI ‘plans to enable the larger window [up to ~1M tokens] and release the model as open source after the trial’ (per TechNode via AI Weekly)
- No Hugging Face repo, license or independent evaluation at launch; vendor showed benchmark screenshots only
- Oct 13–14, 2026: weights published on Hugging Face (inclusionAI/Ling-3.1-flash, MIT license, 1M context); see inclusionAI releases Ling 3.1 Flash open weights
- Predecessors on Vercel: Ling 3.0 Flash (Aug 6, 2026), Ling 3.0 Flash VL (Sept 8) and finance/healthcare variants
What happened
inclusionAI, Ant Group’s foundation-model unit, released Ling 3.1 Flash, a large sparse MoE model built for agentic coding, multi-step analysis and tool-using workflows. It launched as a free two-week trial through third-party gateways rather than as an open-weight release. The company said it would open-source the model after the trial. Exact launch date is uncertain between Sept 29 (Vercel listing) and Sept 30 (TechNode report).
Why it matters
At 560B total parameters it is on the scale of StepFun’s 600B Step 5 Preview, and it follows the same “API or trial first, weights later” pattern is now shared by several Chinese labs. Until the weights and outside evaluations appear, its capabilities are only vendor claims.
Sources
4 sources from 4 sites. Numbers match the chips in the text.
4 sources: 2 primary, 2 press
Primary
- OpenRouter: inclusionAI Ling 3.1 Flashopenrouter.ai, docs
- Vercel AI Gateway: Ling 3.1 Flashvercel.com, docs
Press
- TechNode: Ant Group launches Ling-3.1-flash with 560 billion parameterstechnode.com, press
- AI Weekly: Ant’s InclusionAI ships Ling-3.1-flash with 560B parametersaiweekly.co, press
Changes
- Open-weights release added (key fact, related)
- Filed (found via OpenRouter’s new-model list and AI Weekly).