LLM Models

Google's 2025 flagship reasoning model with a 1M-token context window and strong performance on complex reasoning, coding, and STEM tasks. Fully multimodal (text, image, audio, video). Superseded by the Gemini 3.x generation.

Google's most budget-friendly Gemini 3.1 model (2026), delivering frontier-class capabilities at very low cost for high-volume, latency-sensitive workloads. Fully multimodal with a large context window.

Google DeepMind's advanced reasoning model (February 2026), built for complex problem-solving with a large reasoning boost over Gemini 3 Pro. Fully multimodal with a 1M-token context window; tiered pricing doubles for prompts over 200K tokens.

Google's fast, low-cost Gemini 3 model (December 2025), a major capability upgrade over Gemini 2.5 Flash and a default in the Gemini app. It pairs next-generation intelligence with lightning-fast latency, a ~1M-token context window, and native multimodal input.

DISCONTINUED. Google DeepMind's flagship Gemini 3 model (November 2025), a natively multimodal model with strong reasoning, coding, and agentic performance and a 1M-token context window. It became the top 'Pro' option in the Gemini app and API at launch. Google shut down Gemini 3 Pro Preview on March 9, 2026; the gemini-3-pro-preview alias now points at Gemini 3.1 Pro Preview.

Google's most intelligent model for sustained frontier performance (May 2026), excelling at complex long-horizon agentic and coding tasks. Fully multimodal with a 1M-token context window, combining frontier intelligence with action.

Anthropic's premium Claude model (May 2026), delivering flagship-class reasoning and agentic coding below Fable 5 pricing. Opus 4.8 keeps Opus 4.7's $5/$25 pricing while using ~35% fewer output tokens on agentic coding, with a 1M-token context window, 128K output, adaptive thinking, and mid-conversation system messages.

Anthropic's most agentic Sonnet-class model (June 2026), narrowing the gap to Opus 4.8 on reasoning, tool use, coding, and computer use while staying far cheaper. Features a 1M-token context window with context compaction and adaptive thinking with selectable effort up to 'extra high' (xhigh). It is the default model for Claude Free and Pro plans.

The unsafeguarded sibling of Claude Fable 5 - the same underlying frontier model with safety classifiers lifted, offered only in limited availability to approved Project Glasswing cybersecurity partners. In Anthropic testing, Mythos autonomously found thousands of zero-day vulnerabilities and produced 181 working Firefox exploits on a harness where Opus 4.6 produced two.

Anthropic's first generally available Mythos-class model (June 2026) and its most capable widely released Claude, built for the most demanding reasoning and long-horizon agentic work. Runs with always-on adaptive thinking, a 1M-token context window, and 128K output, and shares Claude Mythos 5's weights with added safety classifiers that route flagged high-risk requests to Opus 4.8.

Anthropic's most capable generally available model (April 2026), with a step-change improvement in agentic coding over Opus 4.6. Features a 1M-token context window at standard pricing, adaptive thinking, and a new tokenizer that contributes to its improved performance (and may use up to ~35% more tokens for the same text).

February 2026 Sonnet release offering the best combination of speed and intelligence, with a 1M-token context window at standard pricing and both extended and adaptive thinking. The recommended balanced model for most production workloads.