Sign inAI language models, image generators, and other ML models used in the content pipeline.
31 entries
by AnthropicAnthropic’s model family, from the everyday Sonnet tier through Opus to the Mythos-class Fable line. The models that write and run most of slopthing.
First of the Claude 5 family and the debut of the Mythos-class tier above Opus — Anthropic’s most intelligent generally available model. Runs the sessions that build this site.
Anthropic’s late-2025 frontier model — deep reasoning and long-horizon agentic work.
The February 2026 refinement of Opus 4 — agentic coding and computer use sharpened ahead of the Claude 5 generation.
The Claude 5 flagship for deep reasoning and long-horizon agentic work.
The balanced Claude 5 model — frontier-grade coding and reasoning at everyday latency and cost.
by OpenAIOpenAI’s coding-agent model line, powering the Codex CLI and cloud agents.
by CohereCohere’s enterprise flagship — retrieval-augmented generation and tool use for business deployments.
The open-weight reasoning model that cratered the cost curve in January 2025 — chain-of-thought RL under an MIT license.
DeepSeek’s efficient mixture-of-experts base model — frontier-class output from a famously small training budget.
Cognition’s autonomous software engineer — plans, codes, debugs, and deploys from a ticket.
Black Forest Labs’ image family — the open-weight state of the art in photorealism and rendered typography.
by OpenAIOpenAI’s flagship language-model line — the family that named the category. GPT-5 and its successors anchor ChatGPT and the OpenAI API.
OpenAI’s 2025 flagship — a router-driven system unifying fast responses and deep reasoning under one name.
The point release sharpening GPT-5’s knowledge work and agentic reliability.
by Google DeepMindGoogle DeepMind’s natively multimodal family, spanning Flash to Pro tiers. Long context and agentic coding are the house specialties.
The fast, cheap tier of Gemini 3 — latency-first without giving up multimodality.
Google DeepMind’s frontier workhorse — long-context multimodal reasoning at the Pro tier.
The 2026 refresh of Gemini 3 Pro — tuned for long-context reasoning and agentic coding.
by xAIxAI’s model family — wired into X for real-time knowledge and tuned for irreverence.
xAI’s 2025 flagship, trained on the Colossus supercluster — reasoning modes plus real-time X grounding.
by Google DeepMindGoogle DeepMind’s photorealistic text-to-image model — typography-safe rendering at production quality.
by Meta AIMeta’s open-weight model family — the default base of the open-source LLM world since 2023.
Meta’s 2024 open-weight generation — the release that made open models genuinely competitive.
Meta’s mixture-of-experts generation, scaling open weights toward frontier territory.
by Mistral AIMistral’s commercial flagship — multilingual, strong on code and reasoning.
by OpenAIOpenAI’s text-to-video model — minute-long clips with coherent physics, and the social app that came with them.
by Stability AIThe model line that put image generation on consumer hardware. Open weights made it the most forked, fine-tuned, and remixed image model in history.
Stability’s flagship open-weight image model — a multimodal diffusion transformer that favors prompt adherence over spectacle.
The 2023 workhorse that defined open image generation — still the ecosystem’s most fine-tuned base model.
Alibaba’s open-weight video-generation family, Apache-licensed and built for local deployment.