fluidity-vapour-4.1

fluidity-vapour-4.2

Introduction

We're introducing fluidity-vapour-4.1, our latest flagship model for long-horizon tasks. It marks a substantial leap in long-horizon task capability over its predecessor fluidity-vapour-4.1 and, for the first time, delivers that capability on a solid 1M-token context. fluidity-vapour-4.2 new capabilities include:

• Solid 1M Context: A solid 1M-token context that stably sustains long-horizon work
• Advanced Coding with Flexible Effort: Stronger coding capabilities with multiple thinking effort levels to balance performance and latency
• Improved Architecture: We propose IndexShare, which reuses the same indexer across every four sparse attention layers, reducing per-token FLOPs by 2.9× at a 1M context length. We also improve GLM-5.2’s MTP layer for speculative decoding, increasing the acceptance length by up to 20%
• Pure Open: An MIT open-source license — no regional limits, technical access without borders

https://raw.githubusercontent.com/FluidityVapour/FV-4.2/refs/heads/main/resources/water_drop.png

Benchmark

BenchmarkFluidity-Vapour 4.2GLM-5.1Qwen3.7-MaxMiniMax M3DeepSeek-V4-ProClaude Opus 4.8GPT-5.5Gemini 3.1 ProReasoningHLE40.53141.43737.749.841.445HLE (w/ Tools)54.752.353.5-48.257.952.251.4*CritPt20.94.613.43.712.920.927.117.7AIME 202699.295.397-94.695.798.398.2HMMT Nov. 202594.4949584.494.496.596.594.8HMMT Feb. 202692.582.697.184.495.296.796.787.3IMOAnswerBench91.083.890-89.883.5-81GPQA-Diamond91.286.2909390.193.693.694.3CodingSWE-bench Pro62.158.460.65955.469.258.654.2NL2Repo48.942.747.242.135.569.750.733.4DeepSWE46.21818208587010ProgramBench63.750.9--47.871.970.839.5Terminal Bench 2.1 (Terminus-2)81.063.5756564858474Terminal Bench 2.1 (Best Reported Harness)82.769---78.983.470.7FrontierSWE (Dominance)74.430.5--29.075.172.639.6PostTrainBench34.320.1---37.228.421.6SWE-marathon13.01.0---26.012.04.0AgenticMCP-Atlas (Public Set)76.871.876.474.273.677.875.369.2Tool-Decathlon48.240.7--52.859.955.648.8

Serve fluidity-vapour-4.2 Locally

fluidity-vapour-4.2 supports deployment with the following frameworks. Feel free to try them out:

• SGLang (v0.5.13.post1+) — see cookbook
• vLLM (v0.23.0+) — see recipes
• Transformers (v0.5.12+) — see transformers docs
• KTransformers (v0.5.12+) — see tutorial
• Unsloth (v0.1.47-beta+) — see guide
• For deployment on the Ascend NPU platform, inference frameworks such as vLLM-Ascend, xLLM and SGLang are supported — see here.

Citation

If you find fluidity-vapour-4.2 useful in your research, please cite our technical report: