ThunderAgent by Together AI: 2x Faster Agentic Inference (ICML 2026)
togethercompute · x · 2026-07-30
Together AI introduced ThunderAgent, a high-throughput system for large-scale agentic inference, accepted as a Spotlight paper at ICML 2026.
The system introduces a novel program-level scheduler abstraction that effectively eliminates KV cache thrashing during multi-turn tool calls.
Key Performance Results:
- More than 2x single-node throughput, with roughly 10x lower P50 latency at high concurrency.
- Delivers a 2.4x speedup on an 8-node cluster, demonstrating near-linear scaling from 16 to 64 GPUs.
Engineering Compatibility:
- Designed as a drop-in solution: requires only a single programid field to plug into existing engine configs (such as KV offloading and speculative decoding).
- Format-agnostic by design, supports OpenAI chat completions today, and has already been adopted by SkyRL and NVIDIA Dynamo.
Related event: Together AI Unveils ThunderAgent, Doubling Agent Inference Throughput(7 posts)→
More from Infra
- NVIDIA launches RTX Spark for local AI on Windows PCs — ostrisai · 2026-10-16
- SkyRL ships Day-0 RL support for Beam with +62% LoRA decode throughput in vLLM — vllm_project · 2026-10-16
- DFlash2 delivers up to 3.03x decode-only speedup for Beam across five benchmarks — vllm_project · 2026-10-16
- vLLM's Beam serving kernels cut GB200 forward time from 8.15 to 6.47 ms at TP8 batch 128 — vllm_project · 2026-10-16
- Reflection's 501B MoE Beam hits 463 tok/s on 8x GB200 with NVFP4 at day 0 — zhyncs42 · 2026-10-16
- Gemma 4 31B QAT on one 24GB card: 46 tok/s plain, ~80 with draft model on RTX 4090 — Subject_Poet_3861 · 2026-10-16