Claude 3.5 Sonnet Review: What It Was, What Happened

Is Claude 3.5 Sonnet still available in 2026?

No — Claude 3.5 Sonnet is a retired Anthropic model, not something you can pick today. Anthropic shipped it in June 2024 and refreshed it that October, then retired both versions from the Claude API on October 28, 2025. It's gone from claude.ai's model picker too, replaced first by the Claude 4 generation and now by Claude Sonnet 5, Anthropic's current mid-tier model as of this writing.

What happened, at a glance

If you're evaluating Claude for a real project today, the model actually worth reading about is whatever Anthropic currently ships as its mid-tier Sonnet — right now that's Claude Sonnet 5. Our Claude AI review covers the current lineup, pricing, and what Claude Code and Artifacts look like today; treat this page as a record of what 3.5 Sonnet was and why it doesn't matter for a 2026 purchase decision.

What Claude 3.5 Sonnet actually was

Claude 3.5 Sonnet was the second-generation Sonnet model in Anthropic's Claude 3 family, sitting in the middle of the original three-tier lineup (Haiku, Sonnet, Opus) but priced and positioned as the default workhorse — Anthropic's own messaging at the time called it faster and cheaper than Claude 3 Opus while beating it on most benchmarks Anthropic published. That's an unusual claim to make about a mid-tier model outperforming your own flagship, and it's part of why 3.5 Sonnet became the model most people associated with "Claude" for well over a year.

The June 2024 release introduced two things that mattered beyond raw benchmark scores. First was a visible jump in coding ability — Anthropic reported a score of 64% on the SWE-bench Verified coding benchmark that it disclosed alongside the release, a figure that helped Claude gain traction with developers who'd previously defaulted to GPT-4-class models for coding assistance. Second was Artifacts, a claude.ai interface feature that let the model generate a piece of content — code, a document, a diagram, a small interactive app — in a dedicated side panel you could keep iterating on, rather than scrolling back through chat history to find the version you wanted. Artifacts is still part of Claude today; 3.5 Sonnet is the model that introduced it.

The October 2024 refresh, which Anthropic released under the same "Claude 3.5 Sonnet" name with a later dated model ID (a naming choice that confused plenty of developers, since there was no version-number bump to signal the change), added computer use in beta — the ability for the model to look at screenshots and control a computer by moving a cursor, clicking, and typing, aimed at automating desktop and browser tasks. Anthropic was explicit at the time that this was an early-stage, error-prone capability meant for developers to experiment with, not a finished product feature, and it shipped through the API rather than as something Free or Pro claude.ai users could turn on directly.

Pricing, while it was active

Claude 3.5 Sonnet's API pricing was $3 per million input tokens and $15 per million output tokens — a number Anthropic held steady across both the June and October releases despite the capability upgrade, which was one of its selling points against rivals that charged more for comparable performance. On claude.ai, it was available on the Free plan (with usage limits) and the paid Pro plan, the same subscription structure Anthropic still uses today, just applied to whichever model is current.

Pricing, free-tier limits, and feature availability change quickly across AI model releases, and none of the figures above should be treated as active pricing — they describe what applied in 2024 and 2025, before retirement. If you're comparing costs today, check Anthropic's current pricing page directly rather than relying on any 2024-era figure, including this one.

Why it was retired

Model retirement isn't unusual — it's standard practice across every major AI lab, and Anthropic publishes a running deprecation schedule for exactly this reason. Anthropic notified developers on August 13, 2025 that both Claude 3.5 Sonnet model IDs (`claude-3-5-sonnet-20240620` and `claude-3-5-sonnet-20241022`) would retire on October 28, 2025, with `claude-sonnet-4-6` named as the recommended replacement at the time — a roughly 11-week runway, longer than Anthropic's 60-day minimum notice commitment.

The practical reason is straightforward: each new Sonnet generation has clearly outperformed the last on the benchmarks Anthropic and independent evaluators track, while Anthropic's infrastructure and safety-testing capacity is finite. Keeping every past model running indefinitely competes for the same compute that trains and serves current ones. Anthropic has said publicly it intends to eventually make retired model weights available again for research purposes, but that hasn't happened for 3.5 Sonnet yet — send a request to either retired model ID today and the API call fails outright.

Where Claude 3.5 Sonnet fit in Anthropic's timeline

• Claude 3 family (March 2024): Haiku, Sonnet, and Opus launch together as three price/performance tiers.
• Claude 3.5 Sonnet (June 2024): A mid-tier model that outperforms the family's own flagship, Claude 3 Opus, on most published benchmarks — an unusual mid-generation leapfrog.
• Claude 3.5 Sonnet, upgraded (October 2024): Same name, new dated model ID, adds computer-use beta.
• Claude 3.7 Sonnet (February 2025): Anthropic's first model with a visible "extended thinking" mode, retired October 2025 in turn.
• Claude 4 family — Opus 4 and Sonnet 4 (May 2025): A full generational jump, with agentic coding as a headline focus.
• Claude Sonnet 4.5, 4.6, and Sonnet 5 (late 2025–2026): Continued point releases refining coding, tool use, and long-context reliability, with Sonnet 5 as Anthropic's current mid-tier model.

That's five Sonnet-line releases in roughly two years — a pace typical of the current AI model market, where "current" has a shelf life measured in months, not years, for any lab shipping competitively.

What actually carried over to today's Claude

Several things people associate with "Claude" started with 3.5 Sonnet and are still core to the product line Anthropic ships now:

• Artifacts — the side-panel content generation feature is still in claude.ai, now extended with more app-like capabilities in later releases.
• The Sonnet tier itself — Anthropic has kept "Sonnet" as its speed/intelligence balance point across every subsequent generation, exactly the positioning 3.5 Sonnet established.
• Computer use, evolved — the beta capability introduced in October 2024 fed directly into the more capable agentic and tool-use features in Claude 4 and later, including what became Claude Code.
• The $3/$15 pricing instinct — later Sonnet generations have shifted pricing (Claude Sonnet 5 lists at $2 input / $10 output per million tokens per Anthropic's current pricing page), but the idea of a mid-tier model priced well below the flagship, aimed at high-volume developer use, traces straight back to how 3.5 Sonnet was positioned.

Pros and cons, as they stood at launch

Where to think twice

If you searched for "Claude 3.5 Sonnet review" hoping to sign up for it today, don't treat any pricing or capability claim above as current — it isn't purchasable, and every number here describes 2024–2025 reality. If you need a model you can deploy right now, skip straight to Anthropic's current docs or our review of the current Claude lineup rather than budgeting around a retired model's numbers. And if your application still has hardcoded calls to `claude-3-5-sonnet-20240620` or `claude-3-5-sonnet-20241022`, those requests are failing in production right now — worth checking today, not after finishing this article.

The bottom line

Claude 3.5 Sonnet mattered. It made a lot of developers take Claude seriously as a default coding assistant instead of a backup when GPT-4 was rate-limited, and Artifacts — a feature it introduced — is still central to how people use Claude today. But it's a 2024-era model Anthropic retired in October 2025, and multiple Sonnet generations (Sonnet 4, then 4.5, 4.6, and now Sonnet 5) have shipped since. There's no legitimate way to call Claude 3.5 Sonnet today — API requests to either dated model ID simply fail. If you're deciding what to build on now, look at Claude Sonnet 5, and check Anthropic's models overview for whatever's newest by the time you're reading this.

FAQ

Can I still use Claude 3.5 Sonnet in 2026?

No. Anthropic retired both dated versions (`claude-3-5-sonnet-20240620` and `claude-3-5-sonnet-20241022`) from the Claude API, Claude Platform on AWS, and Microsoft Foundry on October 28, 2025. Requests to those model IDs now fail, and it's no longer selectable in claude.ai's model picker either.

What replaced Claude 3.5 Sonnet?

Anthropic's official recommended replacement at the time of the retirement notice was `claude-sonnet-4-6`. Since then, Anthropic has shipped Claude Sonnet 5 as its current mid-tier model — check Anthropic's models overview for the exact current lineup, since it changes faster than any single article can track.

Why did Anthropic retire Claude 3.5 Sonnet instead of just leaving it available?

Anthropic maintains a rolling deprecation schedule, giving at least 60 days' notice before retiring any publicly released model. Newer Sonnet generations consistently outperformed 3.5 Sonnet on published benchmarks, and running every past model indefinitely competes for the compute and safety-testing resources needed for current releases.

Was Claude 3.5 Sonnet free to use?

It was available on claude.ai's Free plan (with usage limits) and the paid Pro plan, plus pay-as-you-go API pricing at $3 per million input tokens and $15 per million output tokens. None of that applies today since the model is retired — check Anthropic's current pricing page for what applies now.

Did Claude 3.5 Sonnet support image or video generation?

No. It was a text-and-vision model — it could read and reason about images you uploaded, but couldn't generate images or video. That's still true of Anthropic's current Claude models.

What was "computer use" in Claude 3.5 Sonnet?

A beta capability, added in the October 2024 refresh, that let the model view screenshots of a computer screen and control it — moving the cursor, clicking, typing — to complete tasks through the API. Anthropic described it as an early, error-prone beta, and that line of work fed into the more mature agentic and coding capabilities in later Claude models.

If I have old code calling Claude 3.5 Sonnet, what should I do?

Update your model ID to whatever Anthropic currently lists as active in its models overview, and review its migration guide for behavior differences — later models handle parameters like `temperature` and reasoning controls differently, so a straight model-ID swap without re-testing can change output unexpectedly.

Is there a current Claude model worth reviewing instead?

Yes — see our Claude AI review for what Anthropic's assistant looks like today, including current pricing tiers, Claude Code, and Artifacts. For other AI tool coverage, browse AI & software deals.