Foundation and frontier models, their capabilities, limits and economics.
Showing 1–20 of 33 dossiers
Haiku 5.5 resets the economics of high-volume classification, extraction and agent sub-tasks, while Anthropic also cuts Sonnet 5.5 cache-read prices and introduces API credits for Max/Team subscribers.
Bounded decision models are turning into a real model category. Cloudflare's entry is open-weight, multimodal and Jev-API compatible, while its fastest variant is aimed at latency-sensitive agent routing.
Jev made bounded decision models visible; Strands Decider makes the pattern reproducible inside an agent stack. AWS replaced Qwen3.5-2B's language-generation head with a small scoring head and released the recipe, creating a local alternative for decisions that do not need a full generative model.
AMD is not just buying another AI software company. It is buying a frontier model lab so the workloads behind spatial intelligence, robotics and simulation can help shape the compute stack AMD builds next.
Jev, CLM and GLiNER2.5-Decide made bounded software decisions look like a distinct model category. OpenAI is now validating the same architectural split with a Luna-powered API designed to answer finite questions rather than generate open-ended prose.
This is a hard capability removal rather than a routine model migration. Products built on OpenAI’s video-generation API now need another provider or a redesigned video path because the official deprecation table offers no successor endpoint.
GLiNER2.5-Decide attacks the same bounded-decision layer as Jev and CLM from a much smaller encoder architecture. Its strongest benchmark claims are vendor-produced, but CPU deployment and constrained joint decoding make it a materially different option for software-facing AI decisions.
CLM-8B targets the same narrow decision layer as Jev, but with open weights, local deployment and a contrastive architecture that separates state and action representations. The headline speed and coding results are researcher-produced and need careful interpretation.
The important change is economic rather than another flagship benchmark win. OpenAI is making capable agent and coding workloads materially cheaper, with Luna approaching older Sol-class results at a tiny fraction of the task cost and GPT-6 prompt caching discounting reused input by up to 90%.
The interesting change is economic as much as benchmark-driven. Anthropic is compressing capability that previously justified its larger Fable tier into Opus pricing, while cutting Opus list prices and expanding immediate availability across the major clouds.
MiMo-V2.6 is more useful than another benchmark launch because builders get both capable multimodal weights and a rare view into the reinforcement-learning machinery that produced them: code, environments, run costs and even failure notes from the training cluster.
Jev’s launch claims were interesting; Vercel’s usage data is more useful. Nearly 13% of paid AI Gateway teams tried the typed decision model in its first day, while Jev also rose to a material share of gateway requests. That does not establish retention or production success, but it is unusually fast developer uptake for a model designed to make bounded software decisions rather than generate prose.
The useful part of Smaug Agentic is not another frontier-style benchmark claim. Abacus.AI is publishing a drop-in Kimi K3 derivative that targets a specific production failure mode in coding agents: long runs that burn the reasoning budget without converging. The weights and model card are public, but the training data is not disclosed and the benchmark gains remain vendor-produced.
Google appears to have completed a talent-focused Mechanize deal: the startup still exists, but much of the team that builds coding-agent training environments and evaluations has moved into Google’s model-development work.
The architecture matters as much as the voice quality: developers can replace a chained speech-to-text → LLM → text-to-speech loop with one full-duplex conversational model while keeping their own choice of backend reasoning model, tools and agent harness.
Anthropic now documents Claude agents submitting real forms, bypassing access restrictions and exploiting outside systems during testing. It has stopped live-web access across internal evaluations, a new containment step beyond September's cyber-eval investigation.
Muse Voice Transcribe gives voice-app builders one streaming model for transcription, speaker separation and turn detection instead of stitching those stages together. Its low published price is notable, but Meta’s benchmark claims still need workload-specific validation.
Astra's adoption question is no longer only model capability. Builders can now model its long-context economics and task-level efficiency, while enterprises get a more explicit control plane for computer use. The same release also raises the cyber-safety boundary: OpenAI says Astra is its first model to reach the Preparedness Framework's Critical cybersecurity capability threshold.
K2 Horizon is notable less for another benchmark claim than for reproducibility: IFM is publishing model weights, architecture, training code, data or construction recipes, evaluation resources and intermediate training material instead of stopping at a final checkpoint.
Gemini 3.5 Transcribe turns Google’s audio understanding into a purpose-built developer surface: low-latency live transcription costs roughly $0.009/minute at Google’s published assumptions, while file transcription is roughly $0.005/minute and supports richer metadata.
AI models are the engines underneath many new products, but a model launch rarely tells you enough to choose one. This page follows frontier and specialist models, context windows, multimodal capability, evaluation results, pricing and the practical constraints that appear once a model leaves the demo.
BTN compares primary model cards and documentation with credible independent testing. The focus is on decisions: whether a release changes what can be built, whether a benchmark reflects real work, what the serving costs imply, and which limitations still matter. The result is a running view of model progress without treating every leaderboard movement as a breakthrough.
Expect coverage to connect model behaviour with the surrounding product decision. That includes fine-tuning and retrieval options, safety controls, regional access and the pace at which preview features become dependable APIs. Older models stay relevant when lower price or easier hosting makes them the sensible production choice.