small language models

Dharma-AI released Dharma Field Extractor, a 2-billion-parameter model that turns a document image into typed JSON, either open-ended or matched to a supplied schema. According to Dharma-AI, it scored 0.834 on its 1,638-case benchmark, above every frontier configuration tested, and runs self-hosted on a single L40S.

Liquid AI has released LFM3, a family of six 2B-class models built from LFM3-1.7B-Base, covering text, vision, audio, omni and decision tasks. According to Liquid AI, LFM3-1.7B outperforms LFM2.5-2.6B with 35% fewer parameters and offers a 256K-token context window.

OpenBMB has released MiniCPM-V 4.7 1B, a 1B-parameter multimodal model for single-image, multi-image and video understanding on phones, tablets and local hardware. OpenBMB claims it substantially outperforms Qwen3.5 0.8B in overall score and inference efficiency in Instruct mode. Benchmark figures, context window and training cutoff are not given in the available model card text.

Microsoft has released Decision-1, a decision model built on Qwen3.5-9B for classification, evaluation, and routing. Microsoft claims 83.5% accuracy across 36 benchmarks and 85 ms latency. Input tokens cost $0.042 per million, and output tokens are free.

Microsoft has released FrogNano-4B-2609, a repository-level coding agent derived from Qwen3.5-4B and published under Apache 2.0 with open weights. Microsoft says it was post-trained only with reinforcement learning on about 1,500 synthetic software-engineering tasks, with no stronger-model trajectories. It is evaluated at roughly 131K tokens of context.

Musubi released PolicyLM-1.7B on Tuesday, a 1.7-billion-parameter open-weights decision model built for real-time content moderation. The company claims it applies plain-English policies to messages in under 50 milliseconds and needs no retraining when policies change.

Cactus Compute has released Needle 3, a foundation model compressed to 8-29 MB that runs entirely on-device for tool calling, structured data extraction, and text embedding. The company claims it beats models 10x its size on mobile tool calls while running on hardware as small as microcontrollers.

Inference.net has released Schematron V2 Turbo, a 3-billion-parameter model built specifically for high-volume HTML-to-JSON extraction. The model supports a 128K context window and is priced at $0.03 per 1M input tokens and $0.15 per 1M output tokens.

Inference.net has released Schematron V2 Small, a 3B-parameter model specialized in converting HTML pages into structured JSON output. The model supports a 128K context window and requires extraction schemas to be passed via response_format rather than standard prompts.

Liquid AI has released LFM2.5-VL-3B, a 3.1-billion-parameter vision-language model built for on-device deployment. The model claims leading performance in its size class on grounding, screen understanding, and tool use, while running at 228 tokens/s on an Apple M5 Max.

Nvidia's Nemotron 3.5 Lightning, a 31.6B-parameter open-weight model with only 3.6B active parameters, matches OpenAI's gpt-oss-120b on the Artificial Analysis Intelligence Index while delivering the fastest throughput in its class at nearly 670 tokens per second. The model posts especially large gains on agentic benchmarks, beating both gpt-oss-120b and the larger Nemotron 3 Super.

Liquid AI has released LFM2.5-2.6B, a 2.6B-parameter model designed to run capable tool-calling agents locally on laptops and phones. The company claims it matches or beats models up to 4x its size on instruction-following and tool-use benchmarks while running under 2.5GB of memory.

Liquid AI has released LFM2.5-230M, a 230M parameter hybrid model trained on 19 trillion tokens with a 32,768 token context window. The model achieves 213 tok/s decode speed on Galaxy S25 Ultra and 42 tok/s on Raspberry Pi 5, with support for function calling and data extraction tasks.