Claude Sonnet 5 at $2/$10 per Million Tokens — The Window Closes August 31, 2026
⚠️ Affiliate disclosure — this post contains no affiliate links. Anthropic does not have an affiliate program. All pricing verified from anthropic.com and platform.claude.com as of July 2026.
Claude Sonnet 5 at $2/$10 per Million Tokens — The Window Closes August 31, 2026
Anthropic launched Claude Sonnet 5 on June 30, 2026 — and the introductory pricing is genuinely aggressive. $2 per million input tokens, $10 per million output tokens. That's roughly one-third the cost of Opus 4.8 for performance that's "close to Opus" on agentic coding benchmarks. But the window is short: on September 1, 2026, pricing jumps to $3/$15 — a 50% increase. Here's what Sonnet 5 actually delivers, how it compares to Opus 4.8 and Sonnet 4.6, and why you should lock in your usage before the deadline.
32Days remaining at introductory pricing (as of July 30, 2026)
$2/$10Intro Pricing (per M tokens)
$3/$15Standard Pricing (Sep 1+)
63.2%Agentic Coding Score
~⅓of Opus 4.8 Cost
The Launch
$2/$10Per 1M Tokens
Aug 31Window Closes
200KContext Window
80%Cheaper than Opus
Claude Sonnet 5 is Anthropic's new mid-tier model, launched June 30, 2026, and described by Anthropic as "the most agentic Sonnet yet." It can make plans, use tools like browsers and terminals, and run autonomously at a level that previously required larger, more expensive models. It is now the default model for Claude Free and Pro plans — meaning every free-tier Claude user is already using Sonnet 5.
On the agentic coding benchmark (SWE-bench agentic), Sonnet 5 scored 63.2% — below Opus 4.8's 69.2% but well above Sonnet 4.6's 58.1%. The key number: Sonnet 5 delivers ~91% of Opus 4.8's coding performance at roughly one-third of the cost. For most agentic workloads, that makes Sonnet 5 the pragmatic choice, with Opus 4.8 reserved for the highest-accuracy tasks.
The pricing playAnthropic's introductory pricing is a deliberate strategy to drive adoption of Sonnet 5 as the default agentic model. At $2/$10, it's cheaper than GPT-4.1 for comparable performance. After September 1, at $3/$15 — the same standard price as Sonnet 4.6 — it's still a good deal for a more capable model. But the 50% discount between now and August 31 is the window to maximize value.
Pricing
Complete Claude Model Pricing — July 2026
Model
Input ($/M tok)
Output ($/M tok)
Notes
Sonnet 5INTRO
$2.00
$10.00
Through Aug 31, 2026
Sonnet 5 (standard)
$3.00
$15.00
From Sep 1, 2026
Opus 4.8
$5.00
$25.00
Highest accuracy, most expensive
Fable 5
$10.00
$50.00
Max-tier only, credits for Pro
Haiku 4.5
$1.00
$5.00
Fastest, cheapest Claude
Sonnet 4.6
$3.00
$15.00
Previous Sonnet, being replaced
Pricing verified from platform.claude.com, July 2026. All prices per million tokens. Tokens ≈ ¾ of a word — 1M tokens ≈ 750,000 words.
Analysis
What $2/$10 Actually Buys You
At $2 per million input tokens and $10 per million output tokens, Sonnet 5 is genuinely cheap for agentic work. But agents burn through tokens fast — every turn re-sends the full context: system prompt, prior messages, file contents, tool outputs. A single multi-file refactor can consume hundreds of thousands of input tokens before it's done.
Here's what that looks like in practice:
Light coding session (10 back-and-forth messages, ~50K input tokens, ~5K output tokens): ~$0.15 at intro pricing. Multi-file refactor (30 messages, full file contexts, ~500K input, ~30K output): ~$1.30 at intro pricing. Agentic workflow (autonomous agent, 100+ turns, ~2M input, ~100K output): ~$5.00 at intro pricing. At standard pricing after September 1, those same sessions cost $0.23, $1.95, and $7.50 respectively — a 50% increase.
The agent token trapAgents are token-hungry. Every tool call, every file read, every browser interaction adds to the context window. A single autonomous coding session with Claude Code can easily consume 500K–2M input tokens. At intro pricing ($2/M), that's $1–4. At standard pricing ($3/M), that's $1.50–6. The 50% difference compounds quickly for high-volume agentic workflows. If you're building agent pipelines, lock in your usage before August 31.
Benchmark Comparison
Benchmark
Sonnet 5
Opus 4.8
Sonnet 4.6
Agentic Coding (SWE-bench)
63.2%
69.2%
58.1%
Cost per 1M input tokens
$2 (intro)
$5
$3
Cost per 1M output tokens
$10 (intro)
$25
$15
Performance/cost ratio
Best
Highest quality
Being replaced
Questions
Frequently Asked
01When exactly does the introductory pricing end?
August 31, 2026, at 11:59 PM Pacific Time. On September 1, 2026, pricing moves to $3/M input and $15/M output — the same standard pricing as Sonnet 4.6. Anthropic has been explicit about this date in their launch announcement, pricing page, and API documentation. There is no indication the introductory period will be extended.
02Is Sonnet 5 free for Claude users?
Yes — Sonnet 5 is the default model for Claude Free and Pro plans. If you use Claude via the web app or mobile app on the Free plan, you're already using Sonnet 5. The introductory pricing only applies to API access via the Claude Platform. Pro subscribers ($20/month) get higher rate limits and priority access. Max subscribers ($100/month) get access to Fable 5 and Opus 4.8 as well.
03How does Sonnet 5 compare to Opus 4.8?
Sonnet 5 delivers ~91% of Opus 4.8's agentic coding performance (63.2% vs 69.2% on SWE-bench agentic) at roughly one-third of the cost ($2/$10 vs $5/$25). For most agentic workloads, Sonnet 5 is the pragmatic choice. Opus 4.8 remains the best option for the highest-accuracy tasks where cost is secondary. Anthropic describes Sonnet 5 as "close to Opus 4.8" at a fraction of the price.
04What's the API model name?
Use claude-sonnet-5 in your API calls. For production systems, check Anthropic's documentation for pinned dated model versions if they become available. The migration path from Sonnet 4.6 is straightforward — just change the model string. The API is fully compatible with existing Claude SDK code.
05Should I prepay for credits before the price increase?
Anthropic's API is pay-as-you-go — you can't "lock in" credits at the introductory price. The pricing applies to usage during the introductory period. The best strategy: run your heaviest agentic workloads now, before September 1. If you're building agent pipelines, do your development and testing during the intro period to benefit from the lower rate. After September 1, consider routing lower-priority tasks to Haiku 4.5 ($1/$5) or Sonnet 5 at the standard rate.
06What about Fable 5 — is it worth it?
Fable 5 is Anthropic's most powerful model at $10/M input and $50/M output — 5x the cost of Sonnet 5 at intro pricing. It's permanently available for Max subscribers ($100/month) and available via credits for Pro subscribers ($20/month). For most users, Sonnet 5 at $2/$10 is the better value proposition. Fable 5 is for tasks where accuracy is paramount and cost is no object — legal analysis, complex mathematical reasoning, mission-critical agent decisions.
◆Bottom Line
Final Verdict
Claude Sonnet 5 at $2/$10 per million tokens is the best value in agentic AI right now — but only for 32 more days. It delivers ~91% of Opus 4.8's coding performance at one-third the cost. It's the default model for free Claude users. It's the pragmatic choice for agent pipelines, coding assistants, and autonomous workflows. The introductory pricing ends August 31, 2026 — after that, it's a 50% increase to $3/$15. If you're building agentic systems, run your heaviest workloads now. If you're prototyping, do it during the intro period. If you're on the fence about adopting Claude for agentic tasks, the $2/$10 window is the best reason to start. After September 1, Sonnet 5 at $3/$15 is still a good deal — but it won't be this good again. Get your API key at platform.claude.com and start building before the window closes.
Act Now32 days left at $2/$10 — 50% cheaper than standard pricing. Run heavy agentic workloads before Sep 1.
Best ValueSonnet 5 at intro pricing — ~91% of Opus 4.8 quality at ~⅓ the cost. Default for Free/Pro.
API Modelclaude-sonnet-5 — drop-in replacement for Sonnet 4.6. Compatible with existing Claude SDK.
After Sep 1$3/$15 — same as Sonnet 4.6 standard pricing. Still good value for a more capable model.
