Claude Sonnet 5 benchmarks
11:00 AM · Jun 30, 2026 · 876 Views
Tanishq Mathew Abraham shares benchmarks showing Claude Sonnet 5 beats Sonnet 4.6 across evaluations but trails Opus 4.8 on SWE-bench Pro
Story Overview
Tanishq Mathew Abraham posted benchmark tables that position the just-launched Claude Sonnet 5 as a clear step up from Sonnet 4.6 on agentic coding, terminal use, computer interaction, and multidisciplinary reasoning, while still sitting a few points behind Opus 4.8 on most of the same tests.
9104544.2K
Developer Impact
Gains land near the bigger model
Sonnet 5 reaches within striking distance of Opus 4.8 on SWE-bench Pro, Terminal-Bench, and OSWorld while also posting a slight edge on GPQA-AAA v2, showing how much capability Anthropic squeezed into the lighter tier.
Pricing Watch
Pricing stays Sonnet-level for now
Introductory rates hold at $2 per million input and $10 per million output tokens through August, after which they rise to the usual Sonnet structure, though the new tokenizer may change actual token counts.
Sentiment
Many users criticized Claude Sonnet 5 benchmarks as underwhelming versus Opus 4.8 and GLM models while questioning hype, pricing, and token use, though some praised the specs and fast release cadence.
Pos
32.3%
Neg
67.7%
24 comments with sentiment.
Cluster Engagement
Digg Deeper
No Digg Deeper questions have been answered for this story yet.
Posts from X
Most Activity
Most Activity
VIEWS2KBOOKMARKS1LIKES49
will brown@willccbb
it’s like mythos but if it wasn’t mythos and instead was basically opus 4.7
2hViews 2KLikes 49Bookmarks 1
RETWEETS2REPLIES4
Florian Brand@xeophon
a glm-class model but its three times the price
will brown@willccbb
it’s like mythos but if it wasn’t mythos and instead was basically opus 4.7
2hViews 374Likes 31Bookmarks 1
sankalp@dejavucoder
@willccbb lol
2hViews 470Likes 12
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex
what is the fucking point of saying this for Opus specifically? all compared models are "reference". these jerks are finding new ways to trigger me
Claude@claudeai
Sonnet 5 is a substantial improvement over Sonnet 4.6 on reasoning, tool use, coding, and knowledge work.
Its performance is close to Opus 4.8, at lower prices.
1hViews 381Likes 6Bookmarks 0
Rohan Paul@rohanpaul_ai
On a Agentic search the improvement is so prominent.
Sonnet 4.6 barely improves when you spend more. Sonnet 5, by contrast, gets dramatically better with effort to near-Opus territory on BrowseComp agentic search.
Rohan Paul@rohanpaul_ai
And Claude Sonnet 5 just launched.
Closes the gap with Opus 4.8, and is cheap until August.
This makes agentic AI much cheaper, with $2 input tokens and $10 output tokens per 1M through Aug-26. Price rises after 08-26 to $3 input and $15 output per 1M.
They call Sonnet 5 its “most agentic Sonnet model yet,”
Its coding score hit 63.2% on SWE-bench Pro, versus 58.1% for Sonnet 4.6.
Sonnet 5 gets 63.2% in agentic coding, while Opus 4.8 reaches 69.2% and Sonnet 4.6 hits 58.1%.
But in knowledge work, Sonnet 5 slightly beats Opus 4.8, even though Opus is known for tough judgment and deep research tasks.
1hViews 394Likes 1Bookmarks 1
Andrew Curran@AndrewCurran_
@xeophon @JagersbergKnut My timeline atm.
Florian Brand@xeophon
a glm-class model but its three times the price
1hViews 225Likes 7Bookmarks 0
Florian Brand@xeophon
@henriqueim0veis Everyone regresses on knowledge in favor of agentic web search
1hViews 153Likes 2
henrique imoveis@henriqueim0veis
@xeophon GLM just like any chinese model, has no world knowledge
1hViews 164Likes 1
snow@snowclipsed
@xeophon tvke
1hViews 157Likes 4
Yunfan Zhang@z4y5f3
I think they self-distilled just the right amount so that Sonnet 5 is worse than Opus 4.8 on every benchmark.
will brown@willccbb
it’s like mythos but if it wasn’t mythos and instead was basically opus 4.7
1hViews 814Likes 11Bookmarks 1
Static Snow@StaticSnowman
@teortaxesTex Do you think a human made that graphic?
1hViews 157Likes 3
AntiDisentarian@AntiDisentarian
@teortaxesTex Reference vs direct comparison does seem like a plausible semantic distinction?
1hViews 148Likes 3
Ziwen@ziwenxu_
@xeophon Wild..
1hViews 167Likes 2
xlr8harder@xlr8harder
@teortaxesTex so they could justify highlighting a model without the highest benchmark scores.
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex
what is the fucking point of saying this for Opus specifically? all compared models are "reference". these jerks are finding new ways to trigger me
49mViews 79Likes 2Bookmarks 0
Rafa Schwinger 🇻🇦@Rafa_Schwinger
@willccbb The best feature is spying on you
1hViews 181Likes 1
Nikita Belokopytov@NikiBelokopytov
@teortaxesTex Bro, it's was posted by Sonnet and this is a hallucination
1hViews 137Likes 1
Nirmal Krishnan@nirmal_dist
@xeophon > family guy color palette meme
1hViews 87Likes 1
J A Z I I@notjazii
@xeophon GiB US FABLEEE
1hViews 81Likes 1
Lakshay Sagar Rana@lsrspeakstocomp
@willccbb I am becoming more and more certain that sonnet is a offspring of opus. 5 months for a new model is pretty cool
1hViews 240
MinusGix@MInusGix
@teortaxesTex They want to target people using Sonnet to like their new model, and so 4.6 is "this is an upgrade to it, aren't you Sonnet API users (business) happy?" while Opus is "for reference, looksie, it is nearer to the same tier as our upper tier model but cheaper!!!"
1hViews 75Likes 1
Load more posts