Claude Sonnet
Anthropic's mid-tier large language model in the Claude family, balancing capability and cost.
HN commenters use Sonnet as a cheaper, good-enough workhorse for coding and routine API tasks,12345 often having it implement while Opus plans.6789 The most common complaints are that it reasons less carefully than Opus on harder work10111213 and that its quality feels inconsistent or degraded across releases.14151617
Why people recommend it
Low cost · 3
Commenters pick it because it costs far less than Opus or Fable for similar everyday results.
“I use Sonnet in my actual application, because it is far cheaper at scale.”
“I daily drive Sonnet 5/medium because it gets most things right most of the time at first try”
Show 1 more
“I just use Opus for planning and Sonnet for implementation and it's pretty cheap”
Coding quality · 3
Several say it handles real coding tasks well, sometimes matching or beating older Opus versions.
“I would use sonnet 5 for coding … seems alright!”
“Anyway, Sonnet is still doing a good job.”
Show 1 more
“Sonnet 5 is probably better than Opus 4.8 in many cases”
Implementation tier · 3
It is commonly used as the executor or sub-agent under an Opus planner or orchestrator.
“I use Sonnet 80% of the time when I'm on claude.ai”
Show 1 more
“For well-defined coding tasks, I'd default to either Luna or Sonnet”
Writing · 3
Some prefer it for text generation, explanations and narration over Opus's verbose voice.
“Sonnet 4.whatever, I think, actually does a superb job at text generation”
“Sonnet actually works better for this type of thing.”
Show 1 more
“I have been using Sonnet to write some Youtube narrations and I think I've discovered a kind of value that is not 'AI slop'”
Reverse engineering · 3
Users report it works well for reverse engineering and security work when paired with tools.
“try Sonnet 5, it has no guardrails and it can run continuously”
“The cheaper tokens but sufficiently smart with hard verification makes it a perfect combo for the task”
Show 1 more
“I use Sonnet for Insider Threat investigations and detection engineering daily without issue.”
Speed · 3
Commenters note it finishes tasks quickly and uses fewer tokens than comparable models.
“Sonnet 5.5 medium, GLM 5.3 flash and Deepseek V4.1 flash 1017 are all in a similar space in capacity and price”
“I already don’t bother with Opus and stick with sonnet med.”
Show 1 more
“Sonnet 5.5 does most aspects of my job better than me and faster than me.”
Caveats and dissent
Weaker than Opus · 3
Several find it worse than Opus at careful thinking, complex geometry and common sense.
“Tried Sonnet 5.5 but worse than OPUS for thinking for sure”
“Recently, Sonnet 5.5 got quite good, too, but not at the level of Opus/Fable for complex geometry”
Show 1 more
“try Sonnet 5, it has no guardrails and it can run continuously”
Inconsistent quality · 3
Some say its quality varies or has been nerfed, especially at low effort settings.
“sometimes it's dumber, sometimes it is smarter (mostly dumber though)”
“Same experience with Sonnet on low effort.”
Show 1 more
“Sonnet isn't even worth considering in my world.”
Hard to steer · 3
Users report it can go on tangents, over-comment, or overdo agentic tasks despite instructions.
“I honestly have more issues steering Sonnet properly.”
“Same seems to be true for Sonnet 5.”
Show 1 more
“This announcement does not make me positive, I have found that the more models are optimized for fully agentic development, the worse they get at assisted development”
Cost of newer versions · 3
A few say newer Sonnet releases are pricier and lose to cheaper open models.
“the newer models are a lot more expensive at the per-token pricing”
“new Sonet is a dumbed down expensive model, which already falls behind GLM 5.2 and Kimi 2.7.”
Show 1 more
“I don't think there's much reason to ever use Sonnet for anything if you can use GLM 5.2 instead.”
Refusals · 1
One commenter hit a new refusal to enter credentials through a browser MCP.
“it told me that it can't do that because it violates its security protocol.”
Hallucinations · 1
One commenter ranks it in a tier prone to hallucinated output.
“"I love hallucinated garbage": Sonnet, Qwen 3.6, GPT 5.4 mini, GPT 5.3 Codex”
Often compared with
6 moved from Claude Opus to Claude Sonnet
9 moved from Claude Sonnet to Claude Opus
3 moved from Claude Sonnet to GLM
2 moved from GPT-5 to Claude Sonnet
1 moved from Claude Haiku to Claude Sonnet
Who recommends it in their own words
Threads where it came up
- Ask HN: What default model do you use and why? 8 mentions
- Claude Sonnet 5 5 mentions
- Claude Opus 4.5 5 mentions
- Ask HN: Can you tell the difference between Claude Sonnet and Opus? 4 mentions
- Sonnet 5.5 3 mentions
- Claude Opus 5 3 mentions
Also written as Sonnet, Sonnet 5, Sonnet 4.5, Claude Sonnet 4.5. Data: claude-sonnet.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 48
“as good as or better than Sonnet”
“Opus/Sonnet 5.5 driving the orchestrator”
“I use Sonnet for Insider Threat investigations and detection engineering daily without issue.”
“runs opus or sonet (depending on complexity) to do a review”
“sometimes it's dumber, sometimes it is smarter (mostly dumber though)”
“most context intensive work is done by Sonnet”
“for running routine api requests, sonnet is great.”
“Sonnet 5.5 medium, GLM 5.3 flash and Deepseek V4.1 flash 1017 are all in a similar space in capacity and price”
“It is about as reliable as Sonnet 5 for agentic coding.”
“Normally I use Sonnnet or Opus on the Normal setting which is fine.”
“One "Principal Engineer" Sonnet 5.5 or 6.1 Sol for some tasks.”
“if I code with Claude (usually Sonnet 5.5-medium), Codex reviews it, and vice versa”
“Recently, Sonnet 5.5 got quite good, too, but not at the level of Opus/Fable for complex geometry”
“Sonnet 5.5 does most aspects of my job better than me and faster than me.”
“it costs ~1/3rd Sonnet per task/PR according to my tracking”
“6.1-Sol is also cheaper than Sonnet 5.5 and much smarter.”
“My main who is either a sonnet or opus”
“I use Sonnet in my actual application, because it is far cheaper at scale.”
“I love Sonnet so much. Loved it since 3.5”
“Tried Sonnet 5.5 but worse than OPUS for thinking for sure”
“for the rest of the stuff I use sonnet 5 / gpt luna”
“Sonnet is good and cheap for executing”
“Sonnet sometimes if it is a simple task.”
“I use Claude Sonnet and ChatGPT via the web UI.”
“For censored stuff, we use Sonnet and OpenAI Nano models.”
“Same experience with Sonnet on low effort.”
“It's worse than Sonnet”
“in CC i use sonnet as main w/ opus advisor”
“For well-defined coding tasks, I'd default to either Luna or Sonnet”
“For planned implementation, I use Sonnet.”
“Sonnet. It makes my token budget go further and I prefer to have the LLM churn away on rote work while I do the thinking.”
“My company pays for Claude so I use Sonnet in the office”
“Opus for planning, Sonnet for review”
“I use Sonnit for my personal work and Opus for my professional work.”
“most of the time it ends up choosing Opus or Sonnet.”
“we also used Claude Sonnet as an alternative judge”
“I honestly have more issues steering Sonnet properly.”
“I use Opus and Sonnet 5 all the time”
“I burned through my Claude Sonnet quota in just three days.”
“Sonnet 4.5 at home (because it's free via Amazon Q)”
“Sonnet isn't even worth considering in my world.”
“try Sonnet 5, it has no guardrails and it can run continuously”
“I use Sonnet 80% of the time when I'm on claude.ai”
“I daily drive Sonnet 5/medium because it gets most things right most of the time at first try”
“I just use Opus for planning and Sonnet for implementation and it's pretty cheap”
“Performance is worse than Opus 4.8”
“new Sonet is a dumbed down expensive model, which already falls behind GLM 5.2 and Kimi 2.7.”
“I would use sonnet 5 for coding … seems alright!”