inclusionAI Launches LLaDA2.2-flash: 100B Diffusion Model for Agent Acceleration
inclusionAI released LLaDA2.2-flash, a 100B MoE diffusion language model designed for agentic workloads. Featuring a native 128K context window, it aims to improve multi-turn tool calling and long-context interactions through parallel decoding, offering a novel approach to reducing agent latency.
Confirmed
The model is a 100B MoE with a native 128K context window. Its core mechanism utilizes Levenshtein editing (including KEEP, SUBSTITUTION, DELETE, and INSERT operations) to achieve parallel decoding. In agent evaluations, according to @FellMentKE and @qruiq, the model delivered high throughput on BFCL and SWE-bench, with inference speeds reaching up to 2.3 times that of traditional autoregressive models. @heyshrutimishra also noted a 1.6x overall speedup in agentic workloads.
Unconfirmed
The overall capabilities of diffusion models still face limitations. @Additional-Engine402 pointed out through same-scale comparative evaluations that LLaDA2.2 lags behind autoregressive models in general knowledge and most coding benchmarks, currently holding a speed advantage only in a few agent scenarios.
Why it matters
Diffusion LLMs are expanding into complex agentic applications. @DirectBand896 noted that in multi-turn agent tasks, the true bottleneck slowing down response and driving up inference costs is often the model's token-by-token serial decoding itself, rather than the number of tool calls. By introducing edit-based diffusion and parallel decoding, LLaDA2.2-flash directly addresses this latency pain point, providing a new solution for high-throughput, long-context agent inference.
2026-07-23 ~ 2026-07-25 · 9 related posts
Primary sources
- LLaDA2.2-flash brings 128K context and Levenshtein Editing to agentic diffusion models — pmttyji ·
- Inclusion AI launches LLaDA2.2-flash, a diffusion model for agentic workloads — heyshrutimishra ·
- Side-by-side eval shows diffusion loses overall, but wins speed in agent loops — Additional-Engine402 ·
- [source] LLaDA2.2-flash brings 128K context and Levenshtein Editing to agentic diffusion models — pmttyji · 2026-07-23
- LLaDA2.2-flash brings 100B MoE diffusion model with 128K context and Apache 2.0 — QuixiAI · 2026-07-24
- LLaDA 2.2 claims 703.82 TPS on BFCL-V4 and 519 TPS on SWE-bench Verified — FellMentKE · 2026-07-24
- LLaDA2.2 100B diffusion LLM reportedly matches AR models on agent benchmarks at 2.3× speed — qruiq · 2026-07-25
- LLaDA2.2 targets the real bottleneck in multi-turn agents: decode speed — Direct_Band896 · 2026-07-25
- [source] Side-by-side eval shows diffusion loses overall, but wins speed in agent loops — Additional-Engine402 · 2026-07-25
- Inclusion AI ships LLaDA2.2-flash, a 1.6x faster diffusion model for agents — heyshrutimishra · 2026-07-25
- LLaDA2.2 pushes diffusion LLMs into agentic work with 128K context and self-editing — alifcoder · 2026-07-25
- [source] Inclusion AI launches LLaDA2.2-flash, a diffusion model for agentic workloads — heyshrutimishra · 2026-07-25