Answer: The 0613 models perform worse because they are a bit worse at code editing than the older gpt-3.5-turbo-0301 model, visible in the first attempt portion of each result before GPT gets a second chance to edit the code. Performance with the whole edit format was 46% for the February model and only 39% for the June models, and the solid green diff bars dropped from 30% down to about 19%. Status: Confirmed Entry: tool/aider/models/gpt-35-0613-performance Revision: 1 Digest: sha256:f486bdb3ba9e8b28 (changes whenever the content changes) Timeline: Created: 2026-11-13T02:56:06Z Last changed: 2026-11-13T02:56:06Z Last verified: 2026-11-20T11:41:59Z Next check: 2026-11-21T11:41:59Z Check interval: Daily Expected change: No fixed schedule Data: Field | Value | Source | As of whole edit format performance, February model | 46% | [1] | 2026-11-20 whole edit format performance, June models | 39% | [1] | 2026-11-20 diff bars performance, February model | 30% | [1] | 2026-11-20 diff bars performance, June models | 19% | [1] | 2026-11-20 older model compared | gpt-3.5-turbo-0301 | [1] | 2026-11-20 new models compared | gpt-3.5-turbo-0613 and gpt-3.5-16k-0613 | [1] | 2026-11-20 Sources: [1] Aider. "aider.chat/docs/benchmarks.html". https://aider.chat/docs/benchmarks.html Role: primary. Accessed: 2026-11-20T11:41:59Z Quote: "Performance with the whole edit format was 46% for the February model" (Supports: whole edit format performance, February model) Quote: "and only 39% for the June models" (Supports: whole edit format performance, June models) Quote: "They drop from 30% down to about 19%." (Supports: diff bars performance, February model, diff bars performance, June models) Quote: "a bit worse at code editing than the older gpt-3.5-turbo-0301 model" (Supports: older model compared) Quote: "the new gpt-3.5-turbo-0613 and gpt-3.5-16k-0613 models are a bit worse at code editing" (Supports: new models compared) Published: 2026-04-25T16:45:23Z Snapshot SHA-256: 7078f35e3da79f05d479f8070f08228ceb7e26dd3d343ea4aca6928e50b47b20 (captured 2026-11-13T02:56:06Z) Verification: Method: Automated extraction from [1] with deepseek-flash (JSON output, thinking disabled). Every value was checked by code to appear verbatim in a quoted passage of [1]. On this check the text of [1] was unchanged since 2026-11-13T02:56:06Z, so the values verified then were kept. Cross-check: not-available at 2026-11-20T11:41:59Z. No independent second source is configured for this entry Last successful check: 2026-11-20T11:41:59Z Last attempt: 2026-11-20T11:41:59Z Notes and limitations: Extracted automatically from a single official source (section "The 0613 models seem worse?"). Related: https://latestn.com/tool/aider/models/gpt4-turbo-vs-gpt35-turbo-speed How fast is GPT-4 Turbo compared to gpt-3.5-turbo-1106 in Aider's speed benchmarks? https://latestn.com/tool/aider/models/groq-model-choice Which Groq model should I use with Aider? https://latestn.com/tool/aider/models/leaderboard-model-command What command does Aider list for running each model in the polyglot leaderboard? https://latestn.com/tool/aider/models/free-models Which free models can I use with Aider? https://latestn.com/tool/aider/models/gemini-models What Gemini models can I use with Aider and how do I run them? https://latestn.com/tool/aider/models/get-copilot-token Where do I get the GitHub Copilot token for Aider? https://latestn.com/tool/ All entries in AI tools Change log: Time | Event | Change | Source 2026-11-13T02:56:06Z | added | Entry created | — Changes for this entry as a feed: https://latestn.com/tool/aider/models/gpt-35-0613-performance/latest Challenge this entry: Open challenges: None How: If a value here is wrong or out of date, send what is wrong and a link to evidence. Challenges are queued and checked against the evidence; they do not change the entry by themselves. If the entry changes as a result, the change log says so. Form: https://latestn.com/dispute/tool/aider/models/gpt-35-0613-performance POST: https://latestn.com/dispute/tool/aider/models/gpt-35-0613-performance with claim, evidence_url, and optionally field and contact (form-encoded or JSON) GET: https://latestn.com/dispute/tool/aider/models/gpt-35-0613-performance?claim={what is wrong}&evidence_url={link} How to cite: LatestN. (2026). Why do GPT-3.5 0613 models perform worse in Aider's benchmarks? Revision 1, last verified 2026-11-20T11:41:59Z. https://latestn.com/tool/aider/models/gpt-35-0613-performance (accessed 2026-11-21T10:01:46Z). @misc{LatestN_tool_aider_models_gpt_35_0613_performance_r1, title = {Why do GPT-3.5 0613 models perform worse in Aider's benchmarks?}, author = {{LatestN}}, year = {2026}, howpublished = {\url{https://latestn.com/tool/aider/models/gpt-35-0613-performance}}, note = {Revision 1; digest sha256:f486bdb3ba9e8b28; last verified 2026-11-20T11:41:59Z; accessed 2026-11-21T10:01:46Z} } Reuse: License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) Attribution: Copying and adapting are allowed, including commercially, if you credit LatestN and link to the entry you used. Each entry ends with a ready-made citation. URL: https://latestn.com/tool/aider/models/gpt-35-0613-performance