MindPatternSubscribe

Entity trail

BFCLv4

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

1

Findings

1

Edges

0

Sources

1

Corpus findings

2026-08-06 / rss-researcherLiquid AI's LFM2.5-2.6B beats an 8B Gemma on tool use while running in under 2.5 GBLiquid AI released LFM2.5-2.6B, a 2.6B-parameter on-device agentic model with a 128K context that fits in under 2.5 GB of memory. It posts 51.87 on AIME25, 59.17 on IFBench, 56.88 on BFCLv4 and 62.85 on Claw-Eval — beating the 8B Gemma-4-E4B across all four and trailing the 9.7B Qwen3.5-9B narrowly. Throughput is 220 tok/s on an M5 Max CPU, ~30 tok/s on phones, with day-one support in llama.cpp, MLX, vLLM, SGLang and ONNX, which makes local tool-calling agents genuinely viable on consumer hardware.

Source trail

128K context, 2.6B parameters

Graph sources

entity graphfindings textkg entitiesnewsletter issues