Ordalin
Submit a tool

Reviewed tool profile

Fastino

Open-weight decision models, led by GLiDE, that return probabilities over your own outcomes for agent routing, classification and scoring.

Sources checked

Visit website ↗
Pricing
Pay-as-you-go from prepaid credits. GLiDE inference costs $0.04 per 1M input tokens with free output tokens, and the GLiNER models cost $0.03 per 1M input tokens. Encoder training is billed by GPU time from $0.07 per GPU-minute, and decoder training by token. Official source 1 ↗ · Official source 2 ↗
Available as
Hosted REST API for inference and training; open model weights can also be downloaded and run yourself. Official source 1 ↗ · Official source 2 ↗
Primary group
Automation & Agents

Also consider

See all 6 similar tools ↓

What it does

Fastino is an applied AI lab (Fastino Inc.) that builds open-weight decision and extraction models and sells hosted inference and training through one API. Its flagship is GLiDE, a decision model released on October 1, 2026 that scores application state against caller-defined outcomes and returns probabilities and confidence instead of generated text; the family now spans five open-weight sizes from 0.4B to 27B. The GLiNER models handle entity and relation extraction, and the Training API fine-tunes GLiNER, Nemotron and Qwen models on Fastino's GPUs.

Official source 1 ↗ · Official source 2 ↗ · Official source 3 ↗

Features

  • GLiDE decision model. Evaluates text or structured state against yes/no (Noul), Choice or Score questions and returns an answer with a probability distribution and confidence. Official source 1 ↗ · Official source 2 ↗
  • Decisions instead of prose. Your code defines the allowed outcomes and acts on a bounded result, so there is no generated text to parse. Official source 1 ↗ · Official source 2 ↗
  • Open-weight model family. Five GLiDE decision models from 0.4B to 27B parameters are released as open weights alongside the hosted API. Official source 1 ↗
  • GLiNER extraction and classification. GLiNER 2.5 extracts entities and relations and classifies text; GLiNER2.5-Decide returns typed decisions with confidence scores. Official source 1 ↗
  • Image input. A GLiDE request can include up to four images in its state. Official source 1 ↗
  • Training API. Managed training for GLiNER encoders and interactive training for Nemotron and Qwen decoders, with a Tinker-compatible endpoint for existing Tinker clients. Official source 1 ↗ · Official source 2 ↗

Pricing & free limits

GLiNER inference

$0.03per 1M input tokens

Applies to GLiNER 2.5 and GLiNER-2.5-Decide; output tokens are free. Official source 1 ↗

Encoder training

$0.07per GPU-minute

A100 rate for most GLiNER encoders; GLiNER-2.5-Decide trains on H100 at $0.13 per GPU-minute. Official source 1 ↗

Billing details

Usage is paid from credits that you top up as you go. Official source 1 ↗

Interactive decoder training is billed per token, and checkpoint storage costs $0.05 per GB-month. Official source 1 ↗

Free access & limits

Free access and its limits were not established by the checked official sources.

Best for & limitations

An agent framework asks GLiDE whether a proposed tool call should execute, be revised, need approval or stop. Official source 1 ↗ · Official source 2 ↗

Routing each incoming request to the right model, agent or workflow, with uncertain cases sent to a human. Official source 1 ↗

Support and moderation teams route tickets, moderate content or score urgency. Official source 1 ↗

Checking whether a model's response follows instructions and the required format before it is used. Official source 1 ↗

  • Not a generative model. GLiDE does not write free-form text; open-ended writing, chat, summarization and code generation need a generative model. Official source 1 ↗
  • Single-label questions. Each question returns one label; there is no multi-label primitive. Official source 1 ↗
  • Context and request limits. Each state plus question must fit in 65,536 tokens, and the GLiDE endpoint does not stream, batch independent states or accept fine-tuned deployments. Official source 1 ↗
  • Auto Train is private. The autonomous training agent is in private access and not part of the public API. Official source 1 ↗
Fastino product screenshot
Website previewFixed desktop viewport · 1440 × 900

Similar tools

Published tools that share its categories.

RerunNo-code AI coworkers that run your recurring business tasks on a schedule or trigger, on your own ChatGPT or Claude plan.
Automation & AgentsFree trialChecked Nov 6
Visit
PusharyApprove, deny or answer your AI coding agents' permission requests from your phone or Mac, with per-tool rules and an audit trail.
Automation & AgentsFree trialChecked Nov 6
Visit
CedricAI teammate in Slack, Telegram or Discord that attends meetings and acts across your tools, with approval before anything sends.
Automation & AgentsFree planChecked Nov 5
Visit