Answer: Devin runs the FLUX LoRA training job on its own: it trains the LoRA for about 1,000 steps with validation renders every 100 steps, reports progress to Slack every 15 minutes, picks the best checkpoint (step 700), and deploys it as a serverless Modal endpoint while adding the model to the chat app's model picker. Status: Confirmed Entry: tool/devin/features/fine-tune-flux-lora Revision: 1 Digest: sha256:626ce6e75060e255 (changes whenever the content changes) Timeline: Created: 2026-11-13T03:16:05Z Last changed: 2026-11-13T03:16:05Z Last verified: 2026-11-20T12:48:31Z Next check: 2026-11-21T12:48:31Z Check interval: Daily Expected change: No fixed schedule Data: Field | Value | Unit | Source | As of training steps | about 1,000 steps | steps | [1] | 2026-11-20 validation interval | every 100 steps | steps | [1] | 2026-11-20 Slack progress interval | every 15 minutes | minutes | [1] | 2026-11-20 best checkpoint | step 700 | step | [1] | 2026-11-20 deployment | serverless Modal endpoint | | [1] | 2026-11-20 Sources: [1] Cognition. "Post-Train Custom Models With Devin and Modal". https://docs.devin.ai/tutorial-library/post-training-modal.md Role: primary. Accessed: 2026-11-20T12:48:31Z Quote: "Trains the LoRA**: about 1,000 steps, with validation renders every 100 steps" (Supports: training steps) Quote: "with validation renders every 100 steps" (Supports: validation interval) Quote: "progress every 15 minutes with the latest renders" (Supports: Slack progress interval) Quote: "Picks the best checkpoint**: step 700, since steps 900 and 1,000 showed signs of overfitting." (Supports: best checkpoint) Quote: "serves step 700 as a serverless Modal endpoint" (Supports: deployment) Published: 2026-11-20T12:48:33Z Snapshot SHA-256: e7f25bb09df866fdb44c549ecb7afd0d61b95c59e55ca7be81836c30f32d8edd (captured 2026-11-13T03:16:04Z) 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 2026-11-13T03:16:05Z, so the values verified then were kept. Cross-check: not-available at 2026-11-20T12:48:31Z. No independent second source is configured for this entry Last successful check: 2026-11-20T12:48:31Z Last attempt: 2026-11-20T12:48:31Z Notes and limitations: Extracted automatically from a single official source (section "Step 3: Kick off the training run"). Related: https://latestn.com/tool/devin/features/first-time-prompt-ideas What are some good first-time prompts for Devin? https://latestn.com/tool/devin/features/general-product-features What general product features does Devin offer? https://latestn.com/tool/devin/features/generate-ai-commit-message How do I generate an AI commit message in Devin Desktop? https://latestn.com/tool/devin/features/end-voice-call How do I end a Devin voice call? https://latestn.com/tool/devin/features/existing-engineering-processes How does Devin work within my existing engineering processes and codebases? https://latestn.com/tool/devin/features/fim-support Does Devin Desktop Autocomplete support Fill In The Middle (FIM) suggestions? https://latestn.com/tool/ All entries in AI tools Change log: Time | Event | Change | Source 2026-11-13T03:16:05Z | added | Entry created | — Changes for this entry as a feed: https://latestn.com/tool/devin/features/fine-tune-flux-lora/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/devin/features/fine-tune-flux-lora POST: https://latestn.com/dispute/tool/devin/features/fine-tune-flux-lora with claim, evidence_url, and optionally field and contact (form-encoded or JSON) GET: https://latestn.com/dispute/tool/devin/features/fine-tune-flux-lora?claim={what is wrong}&evidence_url={link} How to cite: LatestN. (2026). How do I fine-tune a FLUX LoRA with Devin on Modal? Revision 1, last verified 2026-11-20T12:48:31Z. https://latestn.com/tool/devin/features/fine-tune-flux-lora (accessed 2026-11-21T10:06:43Z). @misc{LatestN_tool_devin_features_fine_tune_flux_lora_r1, title = {How do I fine-tune a FLUX LoRA with Devin on Modal?}, author = {{LatestN}}, year = {2026}, howpublished = {\url{https://latestn.com/tool/devin/features/fine-tune-flux-lora}}, note = {Revision 1; digest sha256:626ce6e75060e255; last verified 2026-11-20T12:48:31Z; accessed 2026-11-21T10:06:43Z} } 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/devin/features/fine-tune-flux-lora