Wait Which Model?
← Back to directory
Ant Group (InclusionAI)

Ling-3.1-flash

Released Sep 30, 2026 · knowledge cutoff unpublished

Status
Unknown
Location
China
Type
LLM
Modality
Text
Context window
262K
Max output
33K
Speed
—
Price ($/MTok in / out)
— / —
Cost per task
—
Open weights
No
Benchmarks0 of 10 reported

MMLU-Proretired

—

GPQA Diamond

—

SWE-bench Verified

—

SWE-bench Pro

—

Terminal-Bench 2.1

—

AIME (latest)retired

—

Humanity's Last Exam

—

LMArena Elo

—

GDPval-AA v2

—

ARC-AGI-2

—
Strengths & weaknesses

Strengths

  • Hybrid reasoning in one model, so the same endpoint serves quick replies and longer agentic tool-calling runs
  • Pitched by Ant at practical work rather than puzzles — agents, search, office documents and specialist domains such as healthcare
  • Function calling and reasoning come through standard OpenAI-compatible endpoints, so it drops into existing agent stacks

Weaknesses

  • Launched as a two-week free trial with no list price and a 256K cap, so what it will cost and how the promised 1M window behaves are both unknown
  • Open weights are promised but not shipped — nothing to self-host or inspect yet, and no independent evaluation exists
  • Too new to characterise: Ant's three launch scores are all self-reported, with no hands-on reception yet

Announced by Ant Group's Ling (Bailing) team on 2026-09-30 as a roughly 560B-total / 25B-active MoE with up to a 1M-token context; the two-week free trial serves 262,144 tokens (32,768 max output per the Vercel AI Gateway and Novita listings), and Ant says the full window and an open-source release follow the trial, so openWeights is false and licence null until weights publish. Pricing is null rather than $0 because only the free trial exists and no list price is published. Ant reported only GDPval-AA v2.1 Elo 1673, FrontierSWE 75.16 and HealthBench Professional 65.35; none is a tracked key, and gdpvalAA records v2, so the v2.1 figure is kept here rather than in that cell. Tier is fast on Ant's own lineup, where flash sits below the 1T flagship line. Also free inside Command Code (inclusionai/ling-3.1-flash:free) and on Vercel as inclusionai/ling-3.1-flash-free.

For developers

API model strings

  • vercel-ai-gatewayinclusionai/ling-3.1-flash
  • novitainclusionai/ling-3.1-flash

Licence

Proprietary — weights not released

Retirement

No retirement announced

Lineage

No recorded predecessor

News