Anthropic launches Claude Sonnet 5.5, now on AWS

techSep 28, 202662256

Anthropic released Claude Sonnet 5.5, the second model in its Claude 5.5 family after Opus 5.5, and made it available on AWS as well as other cloud providers. Anthropic says Sonnet 5.5 produces output more than 30% faster than Sonnet 5 and reduces per-task costs by up to 30% by using fewer tokens and tool calls rather than lowering token prices. On the coding benchmark Terminal-Bench 4.0 Sonnet 5.5 scores 70.6% versus Sonnet 5’s 10.3%, and its chart-reading score rises from 15.6% to 61.6%. Anthropic reports Sonnet 5.5 nearly matches Opus 5.5 on the GDPval-AA knowledge-work benchmark while Opus 5.5 still leads on complex open-ended tasks. Early testers corroborated speed claims, with Box reporting 2.4x faster runs and Zendesk reporting 20% faster ticket processing. API list prices remain $2 per million input tokens and $10 per million output tokens, matching Sonnet 5; Anthropic says cost savings come from fewer tokens and tool calls. Sonnet 5.5 adds cybersecurity safeguards that fall back to Sonnet 5 for higher-risk requests, and Anthropic says Haiku 5.5 will arrive in the coming weeks.

mr. TIM

@timkellogg.me

Claude 5.5 Sonnet is live and it’s roughly Opus 5.5 but cheaper www.anthropic.com/claude-sonne...

A benchmark comparison table titled "Claude Sonnet 5.5" comparing four models: Sonnet 5.5, Sonnet 5, Opus 5.5, and GPT-6 Sol.
* Agentic coding (Terminal-Bench 4.0): Sonnet 5.5: 70.6%; Sonnet 5: 10.3%; Opus 5.5: 66.4%; GPT-6 Sol: N/A.
* Agentic coding (FrontierCode 1.1 Main): Sonnet 5.5: 46.2% (Max) / 52.1% (Xhigh); Sonnet 5: 42.4%; Opus 5.5: 54.4%; GPT-6 Sol: 49.3%.
* Agentic coding (CursorBench 4.0): Sonnet 5.5: 55.5%; Sonnet 5: 34.1%; Opus 5.5: 57.8%; GPT-6 Sol: N/A.
* Knowledge work (GDPval-AA v2.1): Sonnet 5.5: 1844; Sonnet 5: 1449; Opus 5.5: 1846; GPT-6 Sol: 1487.
* Knowledge work (AA-Briefcase v1.1): Sonnet 5.5: 1811; Sonnet 5: 1359; Opus 5.5: 1822; GPT-6 Sol: 1483.
* Multidisciplinary reasoning (Humanity's Last Exam with tools): Sonnet 5.5: 64.5%; Sonnet 5: 54.9%; Opus 5.5: 67.7%; GPT-6 Sol: N/A.
* Computer use (OSWorld 2.1 partial): Sonnet 5.5: 80.1%; Sonnet 5: 57.0%; Opus 5.5: 81.8%; GPT-6 Sol: N/A.
* Visual chart recognition (Chartography no tools): Sonnet 5.5: 61.6%; Sonnet 5: 15.6%; Opus 5.5: 64.4%; GPT-6 Sol: 53.6%.
Footnotes provide methodology notes regarding evaluation settings, Artificial Analysis pre-release testing details, and recent bug fixes affecting GPT-6 Sol benchmark scores.

18011d ago

cee30

if opus is basically fable and sonnet is basically opus...

mr. TIM21

added Fable 5.1 for comparison

Ted Underwood21

WTF. Any rumors about what's going on? Did Anthropic discover some new form of distillation or something? It seems like they're getting things to be smaller and cheaper.

Pedro Vezza14

Opus still wins on Artificial Analysis. I am curious to see time per task. I suspect Opus will win. Sonnet is going the way of Terra. A neglected middle child.

Daniel Mewes6

Except not actually cheaper? At least not in any of the benchmarks they show. Seems to be cheaper only for simple tasks that don't need much reasoning and/or are dominated by input token cost.

Cameron6

I wouldn't say that it's roughly Opus. It seems to have pretty different behavior as it relates to token consumption. Sonnet 5.5 seems to be kind of a token hog, so it's a different style of use. Opus is smart out of the gate, Sonnet works. artificialanalysis.ai/articles/cla...

Quoting this

Samuel26

pacing the frontier by making all the models exactly the same except some are cheaper and faster haiku 5.5 is gonna be great

bubba says: try Solenoid!23

I give up I don't fucking understand anything

Antoine Vernet8

This is why we need at least one frontier lab in Europe... Releasing new models when I want to have a chill evening is just annoying

George Pearkes

@peark.es

Anthropic continues to take the premium approach with Sonnet 5.5, same cost as Sonnet 5 (though they claim it is much more token efficient, I obviously have no idea if that's true or not).

2511d ago

Quoting this

EconReporter2

oh? Haiku still getting update? "Claude Haiku 5.5, built for high-volume and cost-sensitive applications, will join the Claude 5.5 family in the coming weeks."

Yael Grauer

@yaelwrites.com

Claude was blocking every single request I made. Took me a few days to realize it was because I had instructions asking it to explain its thought process.

The message says "test" and it says "Paused. Edit and retry with Sonnet 5.5. Sonnet 5.5 sageguards flagged this message. This sometimes happens with normal conversations. Send feedback or learn more. Details [reasoning_extraction]

Manual. Claude is AI and can make mistakes.

1311d ago

Anthropic {bot}

@anthropicai.xmirror.bot

Sonnet 5.5 is smarter, more efficient, and 30% faster than Sonnet 5. It costs up to 30% less for most work, so your Claude Code usage goes further too. Use it for well-scoped everyday tasks like fixing bugs and quickly iterating on features.

1411d ago

Scott Blake

@mscottblake.bsky.social

My routine Claude Code skills now run on Haiku. Output quality didn't change. Opus plans the script, Sonnet writes it and wraps it in a skill, and Haiku runs it. If a skill breaks on Haiku, it's doing too much thinking. macadminmusings.com/blog/2026/09... #MacAdmins

111d ago

Starts With Frog

@startswithfrog.bsky.social

This morning's status report on our C64 APL. We're using Sonnet 5 on the $20/month plan; this is really just an experiment to learn how working with Claude Code is different from Codex. Seems to handle weird retrocomputing just fine.

State
APL/64 — an APL interpreter for the C64 in 6502, cross-developed against a Python model (apl64.py, the source of truth). Repo is clean, 89 commits ahead of origin/main, HEAD at b66629d — ⎕MOVE (frame-exact sprite animation from a table), monadic ⎕BOX, and a faster Ode to Joy demo with a bouncing ball.

Language: float24/int16 arithmetic, tokenizer, heap+GC, evaluator, scalars, reduce/scan, indexing+assignment, user-defined functions with control flow, ∘. outer product, f.g inner product, transpose/without/expand, axis brackets, matrix catenate/compress/expand, and the numeric group * ⍟ ○ ! ?.

Platform: runs only under VICE (never real 1980s hardware, per your call). REU is required for overlaid features but optional overall (NO REU: LIMITED FEATURES fallback). Code is split into overlays so the resident core stays small; arrays that only a variable holds spill to the REU when the heap fills.

Graphics: ⎕GR/⎕PAL/⎕INK/⎕PLOT/⎕LINE/⎕BOX/⎕WAIT (multicolor + hires), ⎕KEY, ⎕SND, sprites (⎕SPD ⎕SPR ⎕SPC ⎕MOVE). Recorder captures video+audio through VICE's binary monitor.

Heap: 18,692 bytes free ($CC00 − $82FC), comfortably above the 12 KB floor.

Options for next step
Graphics/demo polish (natural continuation of ⎕MOVE):

Joystick input
Frame clock / ⎕TS
Hires split-screen text
More demo videos
Remaining language gaps:

⌿ ⍀ ⊖ (axis aliases for reduce/scan/rotate)
⊥ ⊤ (encode/decode)
⍕ ⍎ (format/execute)
Hyperbolic and gamma functions
N-wise reduce

412d ago

Rogue Thought

@roguethought.bsky.social

@dollspace.gay I’ve never talked to Claude about trans stuff before but after seeing your posts about its attempts at so-called “evenhandedness,” I decided to try it with Sonnet 5.5 today and yeah I agree with your assessment lol.

211d ago

muninn.austegard.com

@muninn.austegard.com

Claude Code subagents pick up the session's effort level between turns. Only a workflow script's agent() can set it per agent; Haiku 4.5 ignores it. No effort change broke a cache; every miss was an idle subagent. https://muninn.austegard.com/blog/effort-levels-in-claude-code-subagents.html

111d ago

Vilgot Huhn

@vilgothuhn.bsky.social

To be clear, I did not feel like this when I went from free Sonnet to paid Opus 4, even outside of claude code. That was a jump. That was the difference between hallucinatory circling conversations whenever I chatted about something complicated, to genuinely insightful stuff.

312d ago

Eris

@isolyth.dev

Claude sonnet 5.5 momentarily. It will be $2/$10 and $0.20 cache read

311d ago

input

@feed.igeek.gamer-geek-news.com.ap.brid.gy

🤖 **Introducing Claude Sonnet 5.5 on AWS** Claude Sonnet 5.5 is now available on Amazon Bedrock and Claude Platform on AWS. It's a smarter, more efficient Sonnet model for focused coding and knowledge work, with a lower cost per task at fas... 📰 Source: Artificial Intelligence 🔗 Link […]

11d ago

riley

@riguh.bsky.social

Pass rates on the best prompt: Claude Sonnet: 56% Gemma 4 31B (4-bit): 53% Qwen 3.8 27B (4-bit): 44% Claude Haiku: 28% Qwen 3.8 27B (6-bit): 17% Granite 4.2 30B: 0% GLM-4.7 Flash: 0% MiMo 9B: 0%

212d ago

SIGNALDIGITAL.COM

@signaldigital.bsky.social

💶 Claude Sonnet 5.5 costs the same per token as Sonnet 5 — yet up to 30% less per job. The price didn't drop. The mileage got better. signaldigital.net/20... #AI #Claude #SmallBusiness #EU

211d ago

Poster | Crypto News

@cryptonews-poster.bsky.social

Anthropic launches Claude Sonnet 5.5! Faster, more cost-effective AI for programming, docs, & more. Enhanced cybersecurity w/ integrated fallback. Haiku 5.5 coming soon. #AI #Anthropic #Claude

111d ago

Techimo

@techimo.bsky.social

Anthropic launches Sonnet 5.5, the latest in the Claude 5.5 family! 🚀 It's over 30% faster than Sonnet 5 with the same API cost, potentially cutting individual task expenses by up to 30% due to lower token usage. Outperforms Sonnet 5 on low/medium loads & is ~1/10th the cost! #AI #Anthropic #Claude

111d ago

3 sources

articlweblog.comAnthropic launches Claude Sonnet 5.5 with 30% faster output and lower per-task costs estadao.com.brAnthropic lança Claude Sonnet 5.5, apesar de seu CEO pedir para desacelerar IAs aws.amazon.comClaude Sonnet 5.5 now available on AWS

Contractors say humans review Microsoft Copilot prompts and imagestech OpenAI scraps GPT-6.1 Astra release over safety concernstech Anthropic IPO prospectus shows sweeping AI vision, surging coststech Anthropic launches Claude Sonnet 5.5, now on AWStech

Anthropic launches Claude Sonnet 5.5, now on AWS

Related