ling-3.0-tiny

The Opper AI InclusionAI release tracker lists "Ling 3.0 Tiny" as an InclusionAI model released on August 6, 2026, appearing in the August 2026 section alongside Ling-3.0-flash and the later September 2026 Ling-3.0-flash-Fin and Ling-3.0-flash-VL releases. This confirms InclusionAI as the releasing organization and est The tracker also documents the broader InclusionAI/Ling family timeline (Ling 3.0 Tiny, Ling 3.0 Flash, Ling-3.0-flash-VL, Ling-3.0-flash-Fin, plus prior Ling-2.6 and Ring releases), situating Ling 3.0 Tiny within an active release cadence. For the Ling-3.0-tiny subject specifically, the page supplies attribution and a

Ling 3.0 Tiny is a lightweight hybrid reasoning model developed by InclusionAI, Ant Group's AI initiative, according to a third-party technical explainer published on August 17, 2026. The page explicitly names the exact "Ling 3.0 Tiny" variant and describes it as a mixture-of-experts model with roughly 7.9 billion tota The explainer reports a context window of up to 256K tokens, native function calling, a configurable thinking mode, and multiple deployment formats including BF16, FP8, and INT4 weights. Citing model-card figures, it lists an Artificial Analysis Intelligence Index score of 25, an Agentic Index score of 16, output speed

The Artificial Analysis page for Ling 3.0 Tiny confirms it as an InclusionAI open-weights model released in August 2026, with 7.9B total parameters and 1.3B active parameters, a 262k token context window (approximately 393 A4 pages), text-in/text-out modalities, reasoning support (with a noted possible non-reasoning va The page also benchmarks Ling 3.0 Tiny against its specific peer cohort: non-reasoning models versus reasoning models, and open-weights models within the Tiny class (≤4B parameters). Its technical specifications section is model-focused rather than gateway-focused, describing the model's reasoning capability, input/out

The 36Kr/Zhidx article reports that on August 11, 2026, Ant Group's Bailian large model team open-sourced Ling-3.0-tiny, a lightweight hybrid-inference MoE model with 7.9B total parameters and 1.3B activated per token, released in BF16, FP8, and INT4 versions verified on DGX Spark, MacBook, and Mac mini. It cites an Ar The article additionally describes Ling-3.0-tiny's hybrid linear architecture as a 3:1 alternating stack of Moonshot AI's KDA (Kimi Dilation Attention) and DeepSeek's MLA (Multi-head Latent Attention), paired with a sparse MoE feed-forward of 128 routing experts activating 8 routing plus 1 shared expert per token, and

The Baidu encyclopedia entry identifies Ling-3.0-tiny as a native hybrid reasoning model launched by Ant Ling, with a total of 7.9B parameters and only 1.3B activated per token, optimized for mathematical reasoning, instruction following, and resource-sensitive deployment. It states that on August 11, 2026 the Ling Lar The page provides concrete technical and deployment framing for Ling-3.0-tiny that is model-focused rather than gateway-related: it emphasizes the parameter efficiency of activating only 1.3B out of 7.9B per token, and highlights that the model is intended to operate without cloud dependence. Together with its August 1