Every Claude Model
One page, every Claude release. Specs, prices, benchmarks, and when to actually use each model from Claude 3 through Sonnet 4.6.
More than a dozen Claude models have shipped since Anthropic's first release in March 2024. Every generation has brought real capability gains, price resets, or brand-new features. This page collects every model in one place so comparing specs, following the evolution, and picking the right one for your work all happen without jumping around.
Complete Model Timeline
| Date | Model | Key Innovation | Pricing (Input/Output per 1M) | Status |
|---|---|---|---|---|
| Feb 2026 | Sonnet 4.6 | Opus-level coding, 94% computer use | $3/$15 | Active |
| Feb 2026 | Opus 4.6 | 1M context, Agent Teams, PowerPoint | $5/$25 | Active |
| Nov 2025 | Opus 4.5 | 67% cheaper, 76% fewer tokens | $5/$25 | Active |
| Oct 2025 | Haiku 4.5 | Smart model switching, budget tier | Budget | Active |
| Sep 2025 | Sonnet 4.5 | Best coding model, 1M beta context | $3/$15 | Superseded |
| Aug 2025 | Opus 4.1 | Production reliability improvements | $15/$75 | Superseded |
| May 2025 | Claude 4 | Sonnet 4 + Opus 4, Claude Code mainstream | Varies | Active |
| Feb 2025 | Claude 3.7 Sonnet | Hybrid reasoning, extended thinking | $3/$15 | Superseded |
| Oct 2024 | Claude 3.5 Sonnet v2 | Computer Use, upgraded Sonnet + Haiku | $3/$15 | Superseded |
| Jun 2024 | Claude 3.5 Sonnet | Mid-tier beats flagship | $3/$15 | Superseded |
| Mar 2024 | Claude 3 | Three-tier family, vision, 200K context | $0.25-$15 | Retired |
Which Model Should You Use Today?
For most people running Claude Code, the real choice lives between three active models:
Daily driver: Sonnet 4.6. Opus-class smarts at Sonnet prices. Developers pick it over Sonnet 4.5 70% of the time and over Opus 4.5 (November 2025) 59% of the time. At $3/$15 per million tokens, it clears 90%+ of coding tasks without trade-offs.
Heavy lifting: Opus 4.6. Reach for this when the work needs deep reasoning, huge context (1M tokens, now GA with unified pricing), or several agents coordinating at once. Same price as Opus 4.5 ($5/$25), with a real capability bump. Good for complex architecture calls, large codebase work, and agent teams.
Budget tasks: Haiku 4.5. Smart model switching routes the light stuff here automatically. Great for file reads, quick edits, and routine questions. The cheapest tier that still delivers clean output.
If you want a deeper look at tactical model switching inside a session, see the model selection guide. For holding costs down across a team, check the usage optimization guide. Want all of this routing applied out of the box? ClaudeFast's Code Kit ships with routing config for every active Claude model.
How Pricing Evolved
The single most important trend across Claude generations is falling costs. The original Claude 3 Opus came in at $15/$75 per million tokens. Today's Opus 4.6 runs $5/$25 for the same tier, a 67% cut. Capabilities, meanwhile, have shot forward.
| Model | Input (per 1M) | Output (per 1M) | Context Window |
|---|---|---|---|
| Claude 3 Opus | $15 | $75 | 200K |
| Claude 3.5 Sonnet | $3 | $15 | 200K |
| Opus 4.5 | $5 | $25 | 200K (1M beta) |
| Opus 4.6 | $5 | $25 | 1M (GA) |
| Sonnet 4.6 | $3 | $15 | 1M (GA) |
| Sonnet 4.5 | $3 | $15 | 200K (1M beta) |
The direction is plain: every generation hands you more capability per dollar.
Key Milestones
March 2024: The three-tier family with Claude 3. Anthropic locked in the Opus/Sonnet/Haiku naming and rolled out vision. Everything since has been built on top of it.
June 2024: Mid-tier upheaval with Claude 3.5 Sonnet. A cheaper model beating the flagship rewired how developers think about model selection. Bigger stopped meaning better that day.
October 2024: Computer Use with Claude 3.5 Sonnet v2. The first major AI model to steer a computer interface directly. Cursor moves, clicks, typing. A whole new category of agent work.
February 2025: Extended thinking with Claude 3.7 Sonnet. Hybrid reasoning let Claude slow down and think step by step before answering. Quality jumped hard on complex math, science, and multi-step code problems.
May 2025: The Claude 4 generation with Claude 4. Pro-grade coding capabilities turned Claude Code into a daily driver for real development work.
November 2025: The efficiency reset with Opus 4.5. A 67% price drop and 76% fewer output tokens made flagship-class intelligence affordable for everyday use.
February 2026: 1M context and Agent Teams with Opus 4.6. The largest context window on the market (now GA with unified pricing as of March 2026) and native multi-agent coordination opened categories of work that simply were not possible before.
February 2026: Sonnet catches Opus with Sonnet 4.6. For the first time, a Sonnet model was preferred over the previous generation's Opus in coding evaluations. Computer use accuracy hit 94% on insurance benchmarks.
Every model page in this section covers full specs, benchmarks, pricing, and practical guidance for that specific release. Click any model in the timeline above for the full write-up.
Stop configuring. Start building.