FULL STORY

Ant's Ling-3.0-flash: From Launch to Open Source

Ant Group released Ling-3.0-flash, a 124B hybrid reasoning MoE model for agents, later open-sourcing it with only 5.1B active parameters per token.

2026-07-24 ~ 2026-08-06 · 4 episodes · 27 posts

Episode 1 · Ant Group Releases Ling-3.0-flash MoE Model (2026-07-24, 12 posts)

Ant Group's inclusionAI officially released Ling-3.0-flash, a hybrid-reasoning model designed for production-grade agents. The model is now available on OpenRouter and Nous Portal, with OpenRouter offering free access until August 3, 2026, and Nous Portal providing a one-week free trial. Open weights for the model have also been confirmed.

Confirmed

Multiple sources indicate that Ling-3.0-flash is a hybrid-reasoning sparse MoE model. It has a total parameter count of 124B but only activates 5.1B parameters per token, supports a context length of up to 256K, and claims a time-to-first-token (TTFT) of less than 1 second. The model is specifically optimized for programming and agentic workflows. According to @FellMentKE, it features self-correction capabilities, allowing it to read and fix errors autonomously when generated code fails, thereby reducing developer debugging time.

Regarding deployment, Nous Portal announced a one-week free usage period, while OpenRouter's free tier extends directly to August 3, 2026. Furthermore, @victormustar confirmed that open weights for the model are planned.

Why it matters

The combination of a massive total parameter count and highly efficient parameter activation enables the model to deliver strong reasoning capabilities while significantly reducing latency and operational costs. These characteristics make it highly suitable for production-grade agent scenarios that are sensitive to response speed and expenses. The multi-year free access period and upcoming open weights provide developers with substantial convenience and ample room for testing and application.

Episode 2 · Ant Group Releases Ling-3.0-flash Hybrid Reasoning Model (2026-07-27, 2 posts)

Ant Group launched Ling-3.0-flash, a 124B parameter MoE model for production-grade agents that activates only 5.1B parameters per token, with free API access available until August 3rd.

Episode 3 · Ant Group Releases Ling-3.0-flash Hybrid Reasoning Model (2026-07-29, 2 posts)

Ant Group released Ling-3.0-flash, a native hybrid reasoning MoE model for production agents that activates only 5.1B of its 124B parameters per token to rival trillion-parameter flagships.

Episode 4 · Ant Group open-sources Ling-3.0-flash: 124B params, only 5.1B active (2026-08-04, 11 posts)

Ant Group's inclusionAI open-sourced the Ling-3.0-flash model on Hugging Face on August 4. The model has approximately 124B total parameters but activates only 5.1B per token, using a MoE architecture, supporting 256K context, and released under the pure MIT license. The official claims its performance rivals flagship models with 1T activated parameters, and the FP8 quantized version requires only 128GB VRAM for deployment. It is now available on the AI/ML API platform with a free trial period. This combination of extremely low activation and high cost-effectiveness has attracted widespread community attention.

Confirmed

  • Architecture and parameters: Total parameters range from 124B to 127.5B (slight variations across posts), with 5.1B activated parameters (about 1/64 of total). It uses a MoE architecture with 512 routed experts and 1 shared expert, and natively incorporates a hybrid linear architecture from the pretraining stage, supporting 256K context.
  • Deployment cost: Official BF16 and FP8 quantized versions are provided; the FP8 version requires only 128GB VRAM, significantly lowering the hardware barrier for such a large model.
  • Features and benchmark performance: The model supports thinking mode and tool calling, and provides complete chat template configuration. Official reports show strong performance on benchmarks like SWE-bench, and the official claims it can rival flagship models with 1T activated parameters.
  • Open source and API availability: The model is released under the pure MIT license with no commercial restrictions. It is also available on the AI/ML API platform with a free trial period.

Why it matters

  • Ling-3.0-flash balances model capacity and inference cost through extremely low parameter activation, significantly reducing deployment barriers while maintaining high performance, offering the open-source community a highly cost-effective large-scale hybrid inference model.