104B total / 7.4B active (MoE)
256K tokens (262,144)
Unknown
Open-weight instant Ling release focused on high-throughput agentic workloads, long-context document processing, and code-heavy execution. Official materials describe Ling-2.6-flash as a sparse Mixture-of-Experts model with 104B total parameters and 7.4B activated parameters per token, positioned as a cost-efficient successor in the Ling family with full tool-calling and agent collaboration support. Ant's documentation highlights 256K context length, strong long-range retrieval across full-context positions, and production-oriented inference efficiency advantages from low activated parameter count. Third-party launch materials and ecosystem listings describe the model as optimized for token efficiency and agent reinforcement learning, with benchmark claims including strong results on agent-oriented evaluations such as BFCL-V4, TAU2-bench, SWE-bench Verified, Claw-Eval, and PinchBench, while maintaining competitive general intelligence and coding performance for its active compute budget.