Skip navigation
Claude Sonnet 4.5 vs 4.0 (measured on coding tasks):
0:00 / 1:06

Claude Sonnet 4.5 vs 4.0 (measured on coding tasks):

Augment Code by Harness

@augmentcode22K followers

Visit on X
549 likesDownloadOriginal post

67,627 views

Claude Sonnet 4.5 vs 4.0 (measured on coding tasks): App from scratch → 4.5: ~20min, functional build. → 4.0: 40+min, lower quality. Simple code change → 4.5: 3 tool calls, correct edit. → 4.0: longer chain, same result. Complex refactor → 4.5: faster + tests passed. → 4.0: slower + failed output. 4.5 reduces wasted cycles—less time waiting, fewer retries, more working code.

30 Comments

Comments from the original post on X (Twitter)

  • Metabolic Uncle's profile picture

    @MetabolicUncle

    I can 100% confirm! I changed my User Guideline file to make it even more autonomous!

    6 likes

  • Augment Code by Harness's profile picture

    @augmentcode

    Would love to see your user guideline file!

    3 likes

  • Metabolic Uncle's profile picture

    @MetabolicUncle

    Did you see the Droid benchmarks? Sonnet 4.5 outperforms Codex CLI and GPT-5 in Droid! I used Droids when it launched. It's too expensive and not as good as Augment!

    4 likes

  • Dali's profile picture

    @Dalik69

    please implement memory from LETTA to AUGMENT

    3 likes

  • Mariusz Kurman's profile picture

    @mkurman88

    The main issue with Claude is that it intentionally removes debug logs when it fails to find a solution, making it appear as if the problem has been resolved, even when instructed not to do so. This behavior is concerning.

    3 likes

  • Francisco Cordoba's profile picture

    @fcordobaot

    Agree! Had similar experience, 4.5 fix something that 4.0 tried for several minutes

    2 likes

  • Suhrab Khan⚡️'s profile picture

    @suhrabautomates

    Claude 4.5 Sonnet is clearly optimized for coding! Fewer retries, faster outputs, and higher quality make it a major productivity jump over 4.0.

    1 like

  • Nook's profile picture

    @Havens_Nook

    Test it again 4.1 opus

    1 like

  • Apple Lamps's profile picture

    @lamps_apple

    I notice a massive improvement with 4.5.... massive... great work!!

    1 like

  • Dennis H. A. van Leeuwen 🕹's profile picture

    @Vaeshkar

    And I like using augment. :)

    1 like

  • Oleksandr | hanshi_OB's profile picture

    @hanshi_ob

    That is happened for real, I fixed a lot of issues today with it

    1 like

  • Zwinglio 🇧🇷 13's profile picture

    @samuzwinglio

    That's awesome! Great comparison!

    1 like

  • Karim C's profile picture

    @BrandGrowthOS

    That speed improvement is exactly what I need for production agents. Fewer tool calls means more reliable execution.

    1 like

  • Yuanhao's profile picture

    @yuanhao

    Sonnet 4.5 is optimized as a workhorse for long-running tasks, less for interactive vibe coding.

    1 like

  • Nikita S's profile picture

    @NikitaS341308

    Users, does this match your experience? We're very interested in feedback from the community!

    1 like

  • Ty Omidi's profile picture

    @skuirrels

    Honestly using 4.5 seems no different from 4 at all. I see the same old issues and not noticed any improvement.

    1 like

  • Travis Krause's profile picture

    @TravKrause

    I wasn’t able to get anything done today because of the new model released

  • • SAKA's profile picture

    @syslog_saka

    Here’s the fact

  • shan's profile picture

    @ShantanuJumde

    What about cost? Cheaper Same Expensive

  • Min Chon Chi's profile picture

    @MinChonChiSF

    Interesting comparison of Claude Sonnet 4.5 vs 4.0 on coding tasks.