Thinking Machines Lab has released Inkling 2 Small, the newest member of its Inkling model family, and the model scored 31 on version 5.0 of the Artificial Analysis Intelligence Index.

Although the company describes itself as committed to open weights, the new model initially ships as a proprietary product available only through its Tinker platform, with the weights promised for release at a later date.

The launch took place on November 5, 2026, making Inkling 2 Small the lab's second-generation release and arriving less than four months after the original Inkling lineup debuted in July — a fast iteration cycle for a lab that only opened its doors in early 2025.

Architecture and Pricing

Inkling 2 Small is built on a Mixture-of-Experts architecture, combining 594 billion total parameters with only 17 billion active at any given time. Each request is routed to a small subset of the model's internal "experts," allowing the system to hold a large store of knowledge while spending just a fraction of its compute on any single query. That split is why the parameter total reads far larger than the serving bill: per-query compute follows the 17 billion active parameters, not the full 594 billion. Designs like this have become a standard efficiency playbook among today's largest models.

Usage through Thinking Machines' Tinker platform costs $0.30 per 1 million input tokens and $1.00 per 1 million output tokens, pricing the company's research findings describe as competitive with peers.

The model accepts image inputs and supports variable thinking effort, so users can dial up how hard it reasons on a problem, trading speed and cost for depth. Thinking Machines is pitching it for agentic coding, vision, reasoning, and knowledge-work tasks. "Agentic" means the model can carry out multistep jobs on its own rather than simply answering one question at a time.

Training leaned on a large integrated reinforcement-learning run, a process that lets the model use visual feedback to adjust its work in tasks like building websites and games.

Benchmark Results

On benchmarks, Inkling 2 Small ranked 4th overall on DesignArena — a leaderboard focused specifically on design and front-end output — and also placed 1st among open-weights models on that leaderboard. Its AA Intelligence Index score of 31 provides a single-number snapshot of general capability, and because the index is version 5.0, the score reflects the latest edition of that test suite. Since versioned benchmarks reset comparisons with each revision, the 31 is best read alongside other v5.0 scores rather than results from older editions.

Company Background

Thinking Machines Lab launched in February 2025 with a $2 billion seed round at a $12 billion valuation, one of the largest seed raises on record for an AI startup.

The first Inkling model arrived on July 15, 2026. It was a larger system, with 975 billion total parameters and 41 billion active. Inkling-Small followed on July 30, 2026, with 276 billion total parameters and 12 billion active. The new release sits between those two in size.

Inkling 2 Small is more than twice the total size of its namesake predecessor, and its active parameter count rose from 12 billion to 17 billion.

The Open-Weights Question

The most notable tension is the weights question. Thinking Machines frames itself around an open-weights approach, yet Inkling 2 Small currently runs only as a closed model on Tinker.

The company plans to release the weights under an Apache 2.0 license after external safety testing. Apache 2.0 is a permissive license that generally lets anyone use, modify, and build commercial products on the software. For developers, that license marks the difference between renting access through a single vendor and running, or customizing, a model on their own infrastructure.

Until that release happens, developers can only access the model through Thinking Machines' own platform. They can run inference and fine-tune it there, but they cannot download it and host it themselves. No date has been attached to the weights release, which makes the outcome of the external safety testing the main development to watch next.

At $0.30 and $1.00 per million input and output tokens, the model targets teams that want strong coding and design capability without premium-tier bills. The DesignArena results reinforce that pitch: a 4th-place overall finish, combined with the top spot among open-weights models, suggests the visual feedback training is paying off in front-end and design work.

Source: CryptoBriefing — Thinking Machines releases Inkling 2 Small, scores 31 on AI Index. Edited by Diego Almada Lopez.