Claude (Anthropic)
Overview
Claude Series is Anthropic's family of large language models. Founded in 2021 by former OpenAI researchers, Anthropic centers its work around AI safety via the Constitutional AI approach. As of August 2026 the flagship is Claude Opus 5 (July 24, 2026), alongside Sonnet 5 / Fable 5 / Haiku 4.5. Opus 5 ships with a 1M token context window, 128k max output, and adaptive thinking on by default. Claude is widely regarded as having industry-leading coding capabilities.
Latest models: the Claude 5 series
Fable 5.1 and Mythos 5.1 arrived September 1, 2026, cutting Fable's cache-read price 75% (from $1.00 to $0.25 per million tokens; base input/output stay at $10/$50). Fable 5.1 scores 52.6% on Terminal-Bench-Science, ahead of Opus 5's 29.0%.
Lineup
| Model | Use case | API price (input/output per 1M tokens) | Max output |
|---|---|---|---|
| Claude Opus 5 | Flagship: complex agentic coding and enterprise work | $5 / $25 | 128k |
| Claude Sonnet 5 | Balanced, everyday tasks | $2 / $10 | 64k |
| Claude Fable 5.1 | Writing and creative work | — | — |
| Claude Haiku 4.5 | Fastest, low-cost | $1 / $5 | — |
Claude Mythos 5.1 launched alongside Fable 5.1 but remains available only through an approved trusted-access program, not to the general public.
Sonnet 5's $2 / $10 started as introductory pricing set to expire on August 31, 2026, but Anthropic made it permanent on August 10 and cancelled the scheduled increase to $3 / $15.
Opus 5 also offers a fast mode at roughly 2.5x speed ($10 / $50) and the Batch API ($2.50 / $12.50); cache hits cost $0.50 per 1M, a 90% discount on input.
Opus 5 scores 43.3 on Frontier-Bench (Opus 4.8 managed 18.9) and 30.2 on ARC-AGI-3 (versus 1.5), a large jump over the previous generation. Sonnet 5 reaches 63.2% on SWE-bench Pro, close to the 69.2% of the old flagship Opus 4.8.
Key Features
1. 1M Token Context (GA)
Opus 5 and Sonnet 5 include 1M token context at standard pricing. The premium long-context rates have been eliminated.
2. Adaptive Thinking
Claude autonomously decides when and how much to think. At the default effort level (High), Claude almost always thinks, balancing instant answers for simple queries with deep reasoning for complex tasks.
3. Dynamic Web Filtering
Web search and fetch tools now support dynamic filtering — Claude writes and executes code to filter results before they reach the context window.
4. Claude Code (Coding Agent)
Terminal-based autonomous coding agent. Multi-file editing, full-project understanding, and autonomous file operations. Agent Teams enables multi-instance collaboration (experimental).
5. Artifacts / Projects
Real-time preview of code, documents, and apps during chat (Artifacts). Long-term project management with shared context (Projects).
Claude Plans
| Plan | Monthly | Highlights |
|---|---|---|
| Free | $0 | Limited access to Sonnet 5 |
| Pro | $20 | Opus 5, Projects, Artifacts |
| Max (5x) | $100 | 5x Pro usage |
| Max (20x) | $200 | 20x Pro usage, priority access |
| Team / Enterprise | Contact sales | Team features, admin console, SSO |
Strengths
- ✅ Industry-leading coding — top scores on SWE-bench benchmarks
- ✅ 1M token context at standard pricing (GA)
- ✅ Adaptive thinking — autonomous depth-of-thought control
- ✅ Safety-focused — Constitutional AI approach
- ✅ Claude Code — powerful terminal-integrated coding agent
- ✅ Agent Teams — multi-instance collaboration (experimental)
Weaknesses
- ❌ No image/video generation
- ❌ Smaller plugin ecosystem than OpenAI
- ❌ Limited free tier usage
- ❌ Real-time info only via Pro web search
Official Resources
- Website: https://www.anthropic.com
- Claude.ai: https://claude.ai
- API Docs: https://docs.claude.com
- Pricing: https://www.anthropic.com/pricing