Frontier-Model on Best of AI

Claude 3.7 Sonnet

Thu, 16 Jul 2026 00:00:00 +0000

Claude 3.7 Sonnet, released by Anthropic in February 2025, is the company’s first hybrid reasoning model. A single model can answer normally like a standard chat model or switch into an extended thinking mode where it works through a problem step by step before responding, and developers can control how much thinking budget the model uses through the API. This design let Anthropic ship one model that covers both quick everyday questions and harder math, coding, and logic problems, instead of maintaining separate fast and reasoning-focused models. It kept the same 200,000 token context window and $3/$15 per million token pricing as Claude 3.5 Sonnet, and it launched alongside Claude Code, Anthropic’s command-line coding agent, which used 3.7 Sonnet as its default model.

Claude Fable 5

Thu, 16 Jul 2026 00:00:00 +0000

Claude Fable 5 is a public, safety-restricted release of Mythos, an internal Anthropic model class that the company unveiled in April 2026 and initially withheld from general availability because of its advanced cybersecurity capabilities. Anthropic made Fable 5 available to enterprise customers and paid subscribers on June 9, 2026, after adding safeguards that redirect queries in high-risk areas like cybersecurity and biology to Claude Opus 4.8 instead, a mechanism the company says triggers in under 5 percent of sessions. On paper it is the strongest model Anthropic has shipped publicly: 95.0 percent on SWE-bench Verified, 80.3 percent on the harder SWE-bench Pro, and 92.6 percent on GPQA Diamond, with strong results in vision and scientific research tasks as well. It carries a 1 million token context window and is priced at $10 per million input tokens and $50 per million output tokens, double what Anthropic charges for Opus 4.8, reflecting its positioning as a specialty top-tier option rather than a general daily driver.

Claude Opus 4

Thu, 16 Jul 2026 00:00:00 +0000

Claude Opus 4 launched on May 22, 2025 as Anthropic’s flagship model, released alongside the smaller Claude Sonnet 4. It was built for sustained, long-horizon agentic work such as multi-hour coding sessions and research tasks that involve using tools repeatedly across a long thread, rather than answering a single prompt. Anthropic reported it leading on SWE-bench Verified at launch, at roughly 72.5 percent, and it can use tools, including web search, during extended thinking and run several tools in parallel. It runs on a 200,000 token context window, reads images as well as text, and is priced at $15 per million input tokens and $75 per million output tokens, the same rate Anthropic had charged for Claude 3 Opus a year earlier.

Claude Opus 4.1

Thu, 16 Jul 2026 00:00:00 +0000

Claude Opus 4.1, released in August 2025, is a targeted update to Claude Opus 4 rather than a new generation. Anthropic focused the update on coding and agentic accuracy, and it pushed the SWE-bench Verified score up to roughly 74.5 percent, a meaningful jump on the real-world GitHub issue benchmark that Anthropic uses to track coding ability. It kept the same 200,000 token context window, image input support, and $15/$75 per million token pricing as Opus 4, and it gained the ability to use up to 64,000 tokens of extended thinking for harder problems. Anthropic described it as a drop-in replacement for Opus 4, meant to hold the top of the Opus line until the next full model generation shipped.

Claude Opus 4.8

Thu, 16 Jul 2026 00:00:00 +0000

Claude Opus 4.8 is Anthropic’s most capable Opus-tier model, released on May 28, 2026 as an update to Opus 4.7 that fixed some of that model’s comment-verbosity and tool-calling quirks. It scores about 88.6 percent on SWE-bench Verified and 69.2 percent on the harder SWE-bench Pro, ahead of GPT-5.5 and Gemini 3.1 Pro on the same test at the time of release, and Anthropic reports it is more likely to flag its own uncertainty rather than state unsupported claims with confidence. It supports a 1 million token input context window with up to 128,000 tokens of output, and it introduced Dynamic Workflows, a research-preview feature in Claude Code that lets the model plan a large task and then run hundreds of parallel subagents in a single session. Pricing held steady from the previous version at $5 per million input tokens and $25 per million output tokens, with a faster “fast mode” available at $10/$50.

Claude Sonnet 4.5

Thu, 16 Jul 2026 00:00:00 +0000

Claude Sonnet 4.5, released on September 29, 2025, was Anthropic’s highest-scoring Sonnet-tier model on SWE-bench Verified at launch, reaching about 77.2 percent on the full 500-problem set with extended thinking enabled. Anthropic built it heavily around coding and computer use, and testers reported it staying focused on a single complex task for more than 30 hours without losing track of what it was doing. It added context awareness, meaning the model tracks how much of its context window remains during a long conversation and adjusts its behavior as it fills up, which matters for agents running long unattended sessions. It kept Sonnet’s $3/$15 per million token pricing and 200,000 token standard context window, with a 1 million token context window available in beta, making it Anthropic’s primary recommendation for coding agents and long-running automation through the rest of 2025.

Gemini 3.5 Pro

Thu, 16 Jul 2026 00:00:00 +0000

GPT-5.6 Sol

Thu, 16 Jul 2026 00:00:00 +0000

Grok 4.5

Thu, 16 Jul 2026 00:00:00 +0000