GLiNER2.5-Decide (DeBERTa-v3-large, 340M) converted to LiteRT: fp16 weights, answers identical to the official fp32 model on 361/361 test requests, 66–70 ms per request on a Galaxy S26 GPU at 128 tokens, 175 ms at 256. Card, Python and Android sample:

Super fast (and local) browser use, robotics navigation, playing (and beating) LLMs at Minesweeper, and context pruning in Claude Code. Just a few of the GLiNER2.5-Decide demos our team has cooked up. Code and agent skills included. Plug it into Codex, Cursor, Claude Code to download the weights and deploy yourself. If you’ve made an open source demo with GLiNER2.5-Decide, comment here and we’ll put it in our showcase. See all our demos here: Agent skill:

Introducing GLiNER2.5-Decide, our new 340M parameter open weight, encoder-based decision model. GLiNER2.5-Decide is built for fast, deterministic classification. The model evaluates a set of user-defined typed questions and rules, and jointly decodes their answers, returning structured decisions with probability distributions and confidence scores. We evaluated the model’s performance on Fast Decisions, an unseen, internally generated classification suite based on 17 datasets testing real-world use cases across routing, triage, classification, sentiment, and content understanding. Measured against similar decision models, GLiNER2.5-Decide leads in 9 of the 17 datasets, achieving the highest average score: - GLiNER2.5-Decide: 60.1% - SemIf: 56.4% - JevK5: 57.5% - Laya: 46.6% This performance makes the model a strong fit for use cases like tool calling, model routing, browser and computer use, and LLM-as-a-judge. GLiNER2.5-Decide’s lightweight encoder architecture makes it easy to fine-tune the model for specific tasks, while being efficient enough to run locally on consumer-grade CPUs or in air-gapped environments, giving users greater control over where their data goes and where the model runs. To make building and experimenting with GLiNER2.5-Decide as easy as possible, we're also offering hosted inference. You can now use our API to run inference and fine-tune GLiNER models on specialized tasks right inside your own coding agent: As with previous models, we’re also releasing the model weights on @huggingface under the Apache 2.0 license:

We've launched a new GLiNER2 model, heavily tuned for multi-task classification. GLiNER2.5-Decide: 340M encoder, any label set you pass at call time. Intent, routing, priority, sentiment, multi-label tags; several heads scored in one forward pass. No prompt template, no generated tokens. Weights: Fine-tune your own using our agent:

Decision models should do more than pick a label. To truly act, they need to extract evidence, understand relationships, produce structured records, and keep related decisions consistent. And they should do it fast. That’s the new model we built with GLiNER2.5-Decide. Like all GLiNER models, this one is highly responsive to fine-tuning. And now you can fine-tune GLiNER models, including GLiNER2.5-Decide, directly inside your coding agent with our new API. To make building with GLiNER2.5-Decide as fast and simple as possible, we’re also offering hosted inference. Our model goes beyond typed classification. It can return character-level spans, relations, and structured records, then apply implications, exclusions, cardinality limits, and ordinal bounds across related decisions. And it does all of this very fast. On short, single-document requests: CPU latency → 167 ms (48-vCPU Intel Xeon Platinum 8581C) GPU latency → 38 to 47 ms (T4, L4, V100, A100) Run it locally, host it, or fine-tune it. Your decision. Download the weights on @huggingface:

GLiNER2.5 attaches structured judgments to individual spans. A structured judgment is a span attribute: a label from a group you declare, such as sentiment, scored on the mention it describes. In one extract call, the model: - Extracts companies, products, people, or other types you name - Scores that group's labels on each retained span the group applies to - Returns the judgment on that mention, with the span text and offsets One article can praise one company and criticize another. Each company span keeps its own sentiment, tied to the text that describes it, so mixed coverage stays specific to the mention. Watch the full conversation with @urchadeDS and @george_onx to learn more:

One forward pass per move. No generation, no API, no network. GLiNER 2.5 by @fastinoAI running locally on an M4 Pro, playing 5 games concurrently and still able to use my laptop. Best at Sudoku so far. I want to see how many I can run before my fans go into overdrive and my laptop starts levitating Head-to-head against the similar models coming next. Any games that I may have missed ?