InclusionAI

Ant Ling 3.0 Tiny

Name: Ant Ling 3.0 Tiny
Author: InclusionAI

Ant Ling 3.0 Tiny is a InclusionAI model with 131K context, Not verified in / Not verified out per million tokens, last verified October 2, 2026. Open weights: yes. Public-board scores are linked to the publisher; vendor-only figures stay in the watch-out, not on the board.

Ant’s 7.9B open-weight text MoE for local or edge reasoning, code assistance and tool use, with 1.3B active parameters.

Recorded profile verification October 2, 2026Individual provider fields may change

AI model specification details
Specification	Ant Ling 3.0 Tiny
Provider	InclusionAI
Tier	Budget
Context window	131K
Max output	Not verifiedUnverified
Input / 1M tokens	Not verifiedUnverified
Output / 1M tokens	Not verifiedUnverified
Weights	Open
Parameters	7.9B total / 1.3B active (MoE)
Reasoning levels	Not verifiedUnverified
Modalities	text
License	MIT (publisher repository metadata)
API model id	Ling-3.0-tiny
Released	August 7, 2026

Pricing tiers: Downloadable weights declare MIT in publisher metadata; local compute costs apply. The platform lists native API ID Ling-3.0-tiny but its current public pricing table omits this model, so no current per-token tariff is assigned. Native config is 131,072 positions; the exact card’s 262,144 recipe requires YaRN factor 2. A sample 32K generation budget is not a hard output maximum.

Publisher evidence

Published benchmark results

Each result keeps its source date visible. Dates refer to source publication or capture unless the source explicitly states when it measured the result. A missing benchmark is not treated as a zero.

Published benchmark results for Ant Ling 3.0 Tiny
Benchmark	Score	Source date	Source
Artificial Analysis Intelligence IndexLing 3.0 Tiny	11	2026-10-01	Artificial Analysis · View source

Best for

Local text assistants
Small self-hosted tool workflows
Evaluating edge reasoning

Watch out

Thinking is enabled by default and the checkpoint exposes enable_thinking; recommended sampling is temperature=1.0/top_p=0.95/top_k=20. Native API docs restrict their thinking toggle to Flash and graded effort to Ring, so neither is inherited by Tiny. A 1.3B active count is not its full 7.9B weight footprint. Publisher FP8 hardware speeds and memory depend on recipe/context and are not guaranteed for these BF16 weights. Exact weights-publication day, hard output cap and hosted token prices remain unknown. MIT is declared metadata; the attempted raw licence file is missing. AA’s October 1 exact non-estimated Tiny row is an index observation, not a rerun.

History

Ant Ling 3.0 Tiny timeline

Release and status events for this model that passed two-source verification — dated, sourced, and linked.

Ling 3.0 Tiny native-platform launchAvailability updateAugust 7, 2026
Ant Ling official dated platform changelog

Source receipts

Catalog figures for this model were checked against the following sources.

Exact checkpoint: architecture, hybrid thinking, sampling and publisher run settings (accessed 2026-10-02)
Exact native configuration: 131,072 positions, separately extended with YaRN (accessed 2026-10-02)
Publisher repository metadata: MIT declaration and downloadable weight files, creation is not release (accessed 2026-10-02)
First-party platform launch dates; hosted availability is separate from weights (accessed 2026-10-02)
First-party exact API model IDs; named effort controls only apply to Ring (accessed 2026-10-02)
First-party Anthropic contract; video is not listed for this protocol (accessed 2026-10-02)
Exact October 1 Tiny non-estimated public AA index row; capture is not experiment date (accessed 2026-10-02)

Common questions

Ant Ling 3.0 Tiny

Answered from the verified figures on this page rather than general guidance.

What is Ant Ling 3.0 Tiny's context window?

Ant Ling 3.0 Tiny accepts about 131K tokens of context. That only matters if you routinely send very long documents, large codebases, or multi-turn histories that approach that limit.

What is Ant Ling 3.0 Tiny best for?

Ant Ling 3.0 Tiny is a budget tier from InclusionAI. It suits local text assistants, small self-hosted tool workflows, evaluating edge reasoning. Thinking is enabled by default and the checkpoint exposes enable_thinking; recommended sampling is temperature=1.0/top_p=0.95/top_k=20. Native API docs restrict their thinking toggle to Flash and graded effort to Ring, so neither is inherited by Tiny. A 1.3B active count is not its full 7.9B weight footprint. Publisher FP8 hardware speeds and memory depend on recipe/context and are not guaranteed for these BF16 weights. Exact weights-publication day, hard output cap and hosted token prices remain unknown. MIT is declared metadata; the attempted raw licence file is missing. AA’s October 1 exact non-estimated Tiny row is an index observation, not a rerun.

Can I self-host Ant Ling 3.0 Tiny?

Ant Ling 3.0 Tiny publishes open weights, but self-hosting depends on the licence, hardware footprint, quantisation quality, and serving stack. A hosted API is often cheaper until you have measured throughput and concurrency on your own hardware.

Model comparison catalog Benchmark library