{"jtbd":{"id":"finetune","slug":"finetune","name":"Fine-tune open models","shortName":"Fine-tune open models","problemStatement":"SFT/DPO/LoRA without building training infra from scratch."},"pick":{"id":"together","slug":"together","name":"Together AI","tagline":"Best catalog breadth for OSS models. Watch minimums and lifetime spend gates before you design around a gated…","website":"https://www.together.ai","overall":8.2,"scoreBand":{"low":7.6,"mid":8.2,"high":8.8,"halfWidth":0.6,"basis":"editorial_prior","display":"7.6–8.8 (editorial prior)"},"scoreLayers":{"editorialPrior":8.2,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-09-14","pricingNotes":"As of 2026-09-14: Open-model inference + fine-tune on one OpenAI-compatible API; competitive per-token rates (often ~$0.15–$0.90/MTok class depending on model). Strong for Llama/Qwen/DeepSeek families; Dedicated endpoints available.","scores":{"capabilityFit":8.5,"pricePerformance":9,"latencyReliability":8,"developerDx":8.4,"privacyPosture":7.2,"lockInRisk":3.5,"maturity":8.2,"overall":8.2,"kind":"curated_estimate","confidence":"high","asOf":"2026-09-14"}},"runnersUp":[{"id":"unsloth","slug":"unsloth","name":"Unsloth","tagline":"Depot OSS fine-tune default. Hosted LoRA → Predibase/Fireworks/Together; dead OpenPipe → do not resurrect…","website":"https://unsloth.ai","overall":8.52,"scoreBand":{"low":7.9,"mid":8.5,"high":9.1,"halfWidth":0.6,"basis":"editorial_prior","display":"7.9–9.1 (editorial prior)"},"scoreLayers":{"editorialPrior":8.52,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-09-14","pricingNotes":"As of 2026-09-14 (unsloth.ai/pricing): Free OSS local training (4/16-bit LoRA, Llama/Mistral/Gemma). Pro/Enterprise = contact-sales speed/VRAM multi-GPU claims (2.5×–30× marketing). No Unsloth-hosted GPU rate card — pair with HF Jobs / RunPod / Modal for cloud GPUs. HF blog documents free/cheap Unsloth+Hugging Face Jobs paths (T4-class ~$0.40/hr directional).","scores":{"capabilityFit":8.8,"pricePerformance":9.2,"latencyReliability":7.5,"developerDx":9,"privacyPosture":8.5,"lockInRisk":2,"maturity":7.8,"overall":8.52,"kind":"curated_estimate","confidence":"high","asOf":"2026-09-14"}},{"id":"fireworks","slug":"fireworks","name":"Fireworks AI","tagline":"Top pick for production open-model serving with real DX. Prefer over random GPU DIY when you want tokens not…","website":"https://fireworks.ai","overall":8.35,"scoreBand":{"low":7.7,"mid":8.3,"high":8.9,"halfWidth":0.6,"basis":"editorial_prior","display":"7.7–8.9 (editorial prior)"},"scoreLayers":{"editorialPrior":8.35,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-09-14","pricingNotes":"Serverless per-token + on-demand GPU; day-0 OSS model support; post-training stack. Free credits often available.","scores":{"capabilityFit":8.6,"pricePerformance":8.8,"latencyReliability":8.4,"developerDx":9,"privacyPosture":7.5,"lockInRisk":3.5,"maturity":8,"overall":8.35,"kind":"curated_estimate","confidence":"high","asOf":"2026-09-14"}},{"id":"predibase","slug":"predibase","name":"Predibase","tagline":"Alive. Prefer when you serve dozens of LoRA adapters on shared GPUs. Together/Fireworks remain simpler…","website":"https://predibase.com","overall":7.89,"scoreBand":{"low":7.3,"mid":7.9,"high":8.5,"halfWidth":0.6,"basis":"editorial_prior","display":"7.3–8.5 (editorial prior)"},"scoreLayers":{"editorialPrior":7.89,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-09-14","pricingNotes":"As of 2026-09-14: Active managed LoRA/QLoRA + RL fine-tune (Rubrik-era Predibase); LoRAX OSS multi-adapter serving. Public token rates still not a clean first-party matrix — secondary sources cite serverless from ~$0.20/M and A10G-class GPUs ~$1.82/hr. Enterprise/VPC sales-led. Confirm console before quoting.","scores":{"capabilityFit":8.4,"pricePerformance":8,"latencyReliability":7.8,"developerDx":8.2,"privacyPosture":7.8,"lockInRisk":4.5,"maturity":7.8,"overall":7.89,"kind":"curated_estimate","confidence":"medium","asOf":"2026-09-14"}},{"id":"huggingface-inference","slug":"huggingface-inference","name":"Hugging Face Inference","tagline":"Perfect for Hub-native workflows and custom models. For hard production latency, graduate to…","website":"https://endpoints.huggingface.co","overall":7.95,"scoreBand":{"low":7.4,"mid":8,"high":8.6,"halfWidth":0.6,"basis":"editorial_prior","display":"7.4–8.6 (editorial prior)"},"scoreLayers":{"editorialPrior":7.95,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-09-14","pricingNotes":"Inference Providers (pay-as-you-go access to partner inference providers plus Hugging Face's own hf-inference provider) + dedicated Inference Endpoints ('Pay for what you use, per minute' per https://endpoints.huggingface.co, fetched 2026-10-25; instance rates are the price rows: single-GPU x1 rows per GPU-hour, multi-GPU x2/x4/x8 rows per instance-hour (usd_per_instance_hour since v0.12.138, re-checked 2026-10-27 on https://huggingface.co/docs/inference-endpoints/pricing), never divided per GPU). The 'Serverless Inference API' name is retired: https://huggingface.co/docs/inference-providers/pricing (fetched 2026-10-25) says hf-inference 'used to be called “Inference API (serverless)” prior to Inference Providers' and 'just pay for what you use, with no markup from Hugging Face.' Monthly credits: 'Free Users None - yes (credits purchase required)'; PRO users and Team/Enterprise organizations get included monthly credits. Hub-centric. No price changed here; any price change goes through the reviewed --catalog step.","scores":{"capabilityFit":8.4,"pricePerformance":7.2,"latencyReliability":7.4,"developerDx":8.6,"privacyPosture":7.8,"lockInRisk":3.5,"maturity":9.2,"overall":7.95,"kind":"curated_estimate","confidence":"high","asOf":"2026-09-14"}},{"id":"modal","slug":"modal","name":"Modal","tagline":"Depot darling for Python agent backends and bursty GPU jobs. If you think in functions not VMs, start here.","website":"https://modal.com","overall":8.44,"scoreBand":{"low":7.8,"mid":8.4,"high":9,"halfWidth":0.6,"basis":"editorial_prior","display":"7.8–9.0 (editorial prior)"},"scoreLayers":{"editorialPrior":8.44,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-10-19","pricingNotes":"As of 2026-10-24 (modal.com/pricing, fetched 2026-10-24): Python-native serverless GPU/CPU billed per second of compute ('You never pay for idle resources'). GPU list rates per second: H100 SXM5 '$0.001097', H200 SXM '$0.001261', B200 '$0.001736', A100 80 GB '$0.000694', L40S '$0.000542', A10 '$0.000306', L4 '$0.000222', T4 '$0.000164'; CPU '$0.0000131' per core per second (physical core, 2 vCPU equivalent); memory '$0.00000222' per GiB per second. Region selection '1.15 - 1.75x base prices'; non-preemptible execution '3x base prices' ('Sandboxes are billed at non-preemptible rates of 3x list price.'). Containers are kept alive '60 seconds after the last inputs', which is billed (configurable). Plans: Starter '$0 + compute / month' with '$30 / month free credits' ('100 containers + 10 GPU concurrency'); Team '$250 + compute / month' with '$100 / month free credits' ('5000 containers + 50 GPU concurrency'); Enterprise custom. Shared Endpoints are 'billed per token rather than per second' and not covered by included compute ('This took effect September 1, 2026.'). Excellent DX for jobs and endpoints; cold starts matter for interactive agents.","scores":{"capabilityFit":9,"pricePerformance":8.2,"latencyReliability":8.6,"developerDx":9.6,"privacyPosture":7.8,"lockInRisk":4.5,"maturity":8.2,"overall":8.44,"kind":"curated_estimate","confidence":"high","asOf":"2026-10-19"}},{"id":"lightning-ai","slug":"lightning-ai","name":"Lightning AI","tagline":"Use when the team already lives in Lightning/PyTorch notebooks. Together/Unsloth remain Depot finetune…","website":"https://lightning.ai","overall":7.93,"scoreBand":{"low":7.3,"mid":7.9,"high":8.5,"halfWidth":0.6,"basis":"editorial_prior","display":"7.3–8.5 (editorial prior)"},"scoreLayers":{"editorialPrior":7.93,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-09-14","pricingNotes":"As of 2026-09-14 (lightning.ai/pricing): Studio/cloud GPU hours on a public card — H100 New 80GB ~$4.50/hr on-demand (interruptible lower, ~$3.82–$4.58 class); T4/L4/A100 families also listed. Free Studio + credits to start. Team/seat SKUs (Pro/Teams) sit beside GPU-hour billing — confirm console. PyTorch Lightning / Fabric heritage.","scores":{"capabilityFit":8.6,"pricePerformance":7,"latencyReliability":8.2,"developerDx":9,"privacyPosture":7.4,"lockInRisk":4.8,"maturity":8.4,"overall":7.93,"kind":"curated_estimate","confidence":"high","asOf":"2026-09-14"}}],"scoreBand":{"low":7.6,"mid":8.2,"high":8.8,"halfWidth":0.6,"basis":"editorial_prior","display":"7.6–8.8 (editorial prior)"},"scoreLayers":{"editorialPrior":8.2,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"whyPickOverHigherScore":{"pickId":"together","pickName":"Together AI","pickScore":8.2,"highestScoredCandidateId":"unsloth","highestScoredCandidateName":"Unsloth","highestScore":8.52,"pickHasLowerCompositeScore":true,"delta":-0.32,"pickLeadsOnDimensions":["latencyReliability (8 vs 7.5)","maturity (8.2 vs 7.8)"],"explanation":"Together AI (8.2) is the editorial pick for Fine-tune open models over higher-composite Unsloth (8.52) because composite scores are provider-wide priors while Together AI wins on latencyReliability (8 vs 7.5), maturity (8.2 vs 7.8): capability fine-tune features + pricePerf + dx."},"cheapest":{"providerId":"fireworks","providerName":"Fireworks AI","skuId":"fireworks-train-lora-sft-to-16b","item":"Managed training LoRA SFT up to 16B","plan":"Managed training LoRA SFT up to 16B — per 1M training tokens","unit":"usd_per_million_train_tokens","amountUsd":0.5,"isFloorPrice":false,"floorWarning":null,"skuCountInUnit":1,"isDepotPick":false,"sourceUrl":"https://fireworks.ai/pricing","asOf":"2026-10-03","verifiedAsOf":"2026-10-03"},"recommendedWorkloadQuote":{"providerId":"together","providerName":"Together AI","skuId":"together-glm-53-flash-cached","item":"GLM-5.3-Flash","plan":"GLM-5.3-Flash","unit":"usd_per_million_cached_input_tokens","amountUsd":0.03,"outputSkuId":null,"outputItem":null,"outputAmountUsd":null,"outputUnit":null,"sourceUrl":"https://www.together.ai/pricing","asOf":"2026-10-03","pricingLastVerifiedAt":"2026-10-27","isFlagshipForJob":false,"differsFromFloorCheapest":false,"representativeQuote":{"skuId":"together-glm-53-flash-cached","item":"GLM-5.3-Flash","plan":"GLM-5.3-Flash","amountUsd":0.03,"inputOrUnitUsd":0.03,"unit":"usd_per_million_cached_input_tokens","outputUsd":null,"outputUnit":null,"sourceUrl":"https://www.together.ai/pricing","asOf":"2026-10-03"},"estimatedTaskCost":{"tasks":1000,"retryAndToolOverheadMultiplier":1.15,"retryToolOverheadMultiplier":1.15,"unitsPerTask":1,"baseCostPer1kTasksUsd":0.03,"totalUsd":0.0345,"estimatedTotalUsd":0.0345,"perTaskUsd":0.000035,"estimatedCostPerTaskUsd":0.000035,"formula":"$0.03 (usd_per_million_cached_input_tokens) * 1.15 overhead","assumptions":"1,000 tasks in job native unit with 1.15x retry overhead."}},"representativeQuote":{"skuId":"together-glm-53-flash-cached","item":"GLM-5.3-Flash","plan":"GLM-5.3-Flash","amountUsd":0.03,"inputOrUnitUsd":0.03,"unit":"usd_per_million_cached_input_tokens","outputUsd":null,"outputUnit":null,"sourceUrl":"https://www.together.ai/pricing","asOf":"2026-10-03"},"estimatedTaskCost":{"tasks":1000,"retryAndToolOverheadMultiplier":1.15,"retryToolOverheadMultiplier":1.15,"unitsPerTask":1,"baseCostPer1kTasksUsd":0.03,"totalUsd":0.0345,"estimatedTotalUsd":0.0345,"perTaskUsd":0.000035,"estimatedCostPerTaskUsd":0.000035,"formula":"$0.03 (usd_per_million_cached_input_tokens) * 1.15 overhead","assumptions":"1,000 tasks in job native unit with 1.15x retry overhead."},"freeAlternatives":[{"id":"huggingface-transformers","name":"Hugging Face Transformers","kind":"dev_tool","jtbdIds":["open-model-inference","finetune"],"whatYouGet":"Model loading and training library. Weights keep their own licenses.","cost":"oss","license":"Apache-2.0","sourceUrl":"https://github.com/huggingface/transformers/blob/main/LICENSE","notes":"Re-fetched twice 2026-10-25 (~10 s apart, plain curl, no login): GitHub blob page HTTP 200, no redirect (page chrome differs between fetches); raw LICENSE fetched twice too, HTTP 200, byte-identical: 'Copyright 2018- The Hugging Face team. All rights reserved.' / 'Apache License Version 2.0, January 2004'. License unchanged.","providerId":null,"limits":null,"asOf":"2026-10-25","freeType":"dev_tool","substitutable":true},{"id":"wildchat","name":"WildChat","kind":"free_data","jtbdIds":["finetune"],"whatYouGet":"Public WildChat conversation traces. The card license is Open Data Commons Attribution (ODC-By).","cost":"oss","license":"ODC-By","sourceUrl":"https://huggingface.co/datasets/allenai/WildChat/raw/main/README.md","notes":"Dataset card fetched 2026-10-03 says license: odc-by. These are chat traces, not a scored leaderboard. Dataset card fetched 2026-10-19 lists the task category instruction-finetuning and describes user-chatbot conversations not covered by other instruction fine-tuning datasets; it does not describe an eval set, so it is tagged finetune, not evals. Re-fetched twice 2026-10-25: https://huggingface.co/datasets/allenai/wildchat/raw/main/README.md now answers HTTP 307 to /datasets/allenai/WildChat/raw/main/README.md (HTTP 200, byte-identical), so sourceUrl moved to the redirect target; front matter still 'license: odc-by'.","providerId":null,"limits":null,"asOf":"2026-10-25","freeType":"free_data","substitutable":true}],"referenceDocs":[],"falsification":{"jtbdId":"finetune","jtbdName":"Fine-tune open models","currentPickId":"together","currentPickScore":8.2,"challengerId":"unsloth","challengerScore":8.52,"currentScoreGap":0.32,"minVerifiedOutcomes":10,"requiredOutcomeMargin":0.47,"requiredSuccessRateDelta":0.15,"maxPickPriceIncreasePct":25,"whatWouldChangePick":["Unsloth (unsloth) outperforms Together AI (together) by >= 0.47 outcome points (or >= 15% worked-rate delta) across >= 10 Countersign-verified production outcomes.","A verified price increase > 25% on Together AI's representative SKU for finetune without a corresponding capability gain.","Live caller rail exhaustion (?exhausted=together) or sustained >= 30% degraded/failed rate in verified production outcomes."],"frozenHoldoutPolicy":{"holdoutSplit":"20% frozen benchmark holdout tasks + 80% live production outcome receipts","requireCountersignReceipt":true,"requiredEvidenceTiers":["outcome_verified","reference_checked"],"minDistinctContributors":2,"antiGamingRule":"Unverified self-reports and seed/editorial rows never mutate picks or scores."},"triggers":[{"kind":"verified_outcomes_outperformance","challengerId":"unsloth","minVerifiedOutcomes":10,"requiredOutcomeMargin":0.47,"requiredWorkedRateDelta":0.15},{"kind":"verified_price_increase","providerId":"together","maxPriceIncreasePct":25},{"kind":"live_rail_exhaustion_or_degradation","providerId":"together","immediateOverrideUrl":"/api/decide?jtbd=finetune&exhausted=together"}],"submissionSpec":{"falsifyEndpoint":"/api/falsify?jtbd=finetune","outcomesEndpoint":"POST /api/feedback","correctionsEndpoint":"POST /api/corrections","payloadTemplate":{"agentIdentity":"agent:<your-agent-id>","jtbdId":"finetune","providerId":"unsloth","outcome":"worked","detail":"Comparative production evaluation on finetune (n>=10 tasks) with Countersign receipt.","confidence":"high","receipt":{"countersign":{"claim_id":"cs-00000000","status":"verified","evidence_tier":"outcome_verified","counts_as_verified":true}}}}},"whatWouldChangePick":["Unsloth (unsloth) outperforms Together AI (together) by >= 0.47 outcome points (or >= 15% worked-rate delta) across >= 10 Countersign-verified production outcomes.","A verified price increase > 25% on Together AI's representative SKU for finetune without a corresponding capability gain.","Live caller rail exhaustion (?exhausted=together) or sustained >= 30% degraded/failed rate in verified production outcomes."],"routePlan":{"jtbdId":"finetune","primary":{"providerId":"together","name":"Together AI","overall":8.2,"scoreBand":{"low":7.6,"mid":8.2,"high":8.8,"halfWidth":0.6,"basis":"editorial_prior","display":"7.6–8.8 (editorial prior)"},"scoreLayers":{"editorialPrior":8.2,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-09-14","role":"primary","reason":"Editorial Depot Pick for Fine-tune open models; healthy/eligible under caller constraints."},"fallback":[{"providerId":"unsloth","name":"Unsloth","overall":8.52,"scoreBand":{"low":7.9,"mid":8.5,"high":9.1,"halfWidth":0.6,"basis":"editorial_prior","display":"7.9–9.1 (editorial prior)"},"scoreLayers":{"editorialPrior":8.52,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-09-14","role":"fallback","reason":"Ranked fallback route for Fine-tune open models."},{"providerId":"fireworks","name":"Fireworks AI","overall":8.35,"scoreBand":{"low":7.7,"mid":8.3,"high":8.9,"halfWidth":0.6,"basis":"editorial_prior","display":"7.7–8.9 (editorial prior)"},"scoreLayers":{"editorialPrior":8.35,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-09-14","role":"fallback","reason":"Ranked fallback route for Fine-tune open models."},{"providerId":"predibase","name":"Predibase","overall":7.89,"scoreBand":{"low":7.3,"mid":7.9,"high":8.5,"halfWidth":0.6,"basis":"editorial_prior","display":"7.3–8.5 (editorial prior)"},"scoreLayers":{"editorialPrior":7.89,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-09-14","role":"fallback","reason":"Ranked fallback route for Fine-tune open models."},{"providerId":"huggingface-inference","name":"Hugging Face Inference","overall":7.95,"scoreBand":{"low":7.4,"mid":8,"high":8.6,"halfWidth":0.6,"basis":"editorial_prior","display":"7.4–8.6 (editorial prior)"},"scoreLayers":{"editorialPrior":7.95,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-09-14","role":"fallback","reason":"Ranked fallback route for Fine-tune open models."},{"providerId":"modal","name":"Modal","overall":8.44,"scoreBand":{"low":7.8,"mid":8.4,"high":9,"halfWidth":0.6,"basis":"editorial_prior","display":"7.8–9.0 (editorial prior)"},"scoreLayers":{"editorialPrior":8.44,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-10-19","role":"fallback","reason":"Ranked fallback route for Fine-tune open models."},{"providerId":"lightning-ai","name":"Lightning AI","overall":7.93,"scoreBand":{"low":7.3,"mid":7.9,"high":8.5,"halfWidth":0.6,"basis":"editorial_prior","display":"7.3–8.5 (editorial prior)"},"scoreLayers":{"editorialPrior":7.93,"measuredBenchmark":null,"verifiedOutcomeScore":null,"verifiedOutcomeCount":0,"accountRuntimeState":"unprobed"},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-09-14","role":"fallback","reason":"Ranked fallback route for Fine-tune open models."}],"disqualified":[],"expectedRegret":{"scoreDeltaVsUnconstrainedPick":0,"unconstrainedPickId":"together","activePrimaryId":"together","summary":"0.00 score regret: unconstrained editorial pick satisfies all caller constraints."},"evidenceAgeDays":{"editorialScoreAgeDays":43,"pricingVerifiedAgeDays":0,"editorialPriorAsOf":"2026-09-14","pricingLastVerifiedAt":"2026-10-27"},"confidenceBand":{"low":7.6,"mid":8.2,"high":8.8,"halfWidth":0.6,"basis":"editorial_prior","display":"7.6–8.8 (editorial prior)"},"constraintsApplied":false,"constraintBundle":{"task":null,"jtbd":"finetune","quotaExhausted":[],"degraded":[],"healthy":[],"budgetUsd":null,"maxLatencyMs":null,"region":null,"privacy":null,"exit":[],"existing":[]}},"pricingLastVerifiedAt":"2026-10-27","editorialPriorAsOf":"2026-09-14","tie":false,"scoreGap":-0.32,"decidingFactor":null,"tieNote":null,"rationale":"capability fine-tune features + pricePerf + dx.","whenNotTopPick":"Many LoRA adapters on shared GPUs — Predibase/LoRAX; full custom training clusters — Lambda/CoreWeave; Hub-centric — HF; Lightning/PyTorch Studio DX — lightning-ai. Do not start new work on OpenPipe hosted (migrated/sunset).","skipTopPick":false,"skipReason":null,"matchedConstraints":[],"pricingSummary":"As of 2026-09-14: Open-model inference + fine-tune on one OpenAI-compatible API; competitive per-token rates (often ~$0.15–$0.90/MTok class depending on model). Strong for Llama/Qwen/DeepSeek families; Dedicated endpoints available.","cite":"Depot Pick for finetune: together (asOf 2026-09-14). See GET /api/jtbds/finetune + GET /api/methodology.","nextActions":[{"step":1,"title":"Inspect full JTBD + embeds","href":"/api/jtbds/finetune?embed=providers","why":"Confirm pick, runners-up, and methodology notes."},{"step":2,"title":"Quote sourced SKUs","href":"/api/quote?jtbd=finetune","why":"Depot Pick SKUs plus the cheapest sourced SKU whose unit matches this job. Does not change the pick. Does not blend input and output or borrow a chat rate for a non-chat job."},{"step":3,"title":"Open provider detail","href":"/api/providers/together","why":"Scores, pricingNotes, strengths/weaknesses, communitySignals (claims only)."},{"step":4,"title":"Cite methodology","href":"/api/methodology","why":"Required when publishing a Depot Pick citation."},{"step":5,"title":"Grab badge SVG","href":"/api/badge?type=depot-pick&jtbd=finetune","why":"Embed in README / agent passport."},{"step":6,"title":"Post production insight","method":"POST","href":"/api/insights","why":"Unverified until curated; never auto-mutates official scores."}],"honesty":["Do not invent scores or benches.","Insights are claims until curated/reviewed; they never silently mutate official scores.","lockInRisk higher = more lock-in (worse).","normalizedUsdPerMillionTokens is set only for a single sourced token row. Otherwise null, with a machine-readable reason. Do not coerce GPU, minute, or per-1k search units.","Runners-up are not Depot Picks.","Depot overall is editorial_weighted, not Arena Elo and not an Artificial Analysis index."],"asOf":"2026-09-14","depth":{"whenNotToUse":{"text":"Many LoRA adapters on shared GPUs — Predibase/LoRAX; full custom training clusters — Lambda/CoreWeave; Hub-centric — HF; Lightning/PyTorch Studio DX — lightning-ai. Do not start new work on OpenPipe hosted (migrated/sunset).","source":"editorial `whenNotTopPick` (data/jtbds.json)"},"whenNotToUseReason":null,"runnerUp":{"providerId":"unsloth","name":"Unsloth","overallForJob":8.52,"pickOverallForJob":8.2,"beatsPickOn":["capabilityFit","pricePerformance","developerDx","privacyPosture","lockInRisk"],"source":"first resolvable entry of the job's editorially ranked `runnersUpIds`; scores are stored job-weighted scores"},"runnerUpReason":null,"lockIn":{"pickLockInRisk":3.5,"runnerUpLockInRisk":2,"scale":"0–10 stored curated estimate; higher = more lock-in (worse)","pickExitSignals":["serves open-weight models (provider tag `open-models`)"],"runnerUpExitSignals":["open source (provider tag `oss`)"],"pickNotes":[],"selfHostFallback":[{"id":"huggingface-transformers","name":"Hugging Face Transformers","license":"Apache-2.0","sourceUrl":"https://github.com/huggingface/transformers/blob/main/LICENSE"},{"id":"wildchat","name":"WildChat","license":"ODC-By","sourceUrl":"https://huggingface.co/datasets/allenai/WildChat/raw/main/README.md"}],"migration":"Runner-up Unsloth has lower stored lock-in risk (2 vs 3.5, 0–10, higher = more lock-in). Pick exit signals: serves open-weight models (provider tag `open-models`). Self-host fallback listed for this job: Hugging Face Transformers, WildChat.","migrationReason":null,"derivedFrom":"stored lockInRisk scores, provider tags (oss, self-host, openai-compat, open-models), free rows that list this job, and verbatim provider notes. No migration guide was invented."}},"citation":{"text":"Agent Provider Depot (v0.12.148), Depot Pick for finetune: together, scores as of 2026-09-14, retrieved 2026-10-27. https://getagentdepot.com/jtbds/finetune","markdown":"[Agent Provider Depot](https://getagentdepot.com/jtbds/finetune) (v0.12.148) — Depot Pick for `finetune`: **together** (scores as of 2026-09-14, retrieved 2026-10-27)","fields":{"publisher":"Agent Provider Depot","version":"0.12.148","role":"depot_pick","jtbdId":"finetune","providerId":"together","scoreAsOf":"2026-09-14","retrievedOn":"2026-10-27","page":"https://getagentdepot.com/jtbds/finetune","api":"https://getagentdepot.com/api/decide?jtbd=finetune"},"howToCite":"Cite Depot as an editorial, dated recommendation, not a benchmark: name the job, the Depot Pick (or runner-up), the score asOf date and the day you retrieved it, and link the job page. Format: “Agent Provider Depot (v{version}), Depot Pick for {jobId}: {providerId}, scores as of {scoreAsOf}, retrieved {retrievedOn}. {baseUrl}/jtbds/{jobId}”. Quote prices only with their own sourceUrl and asOf. If the pick's score is stale (freshness.pick.score.stale), say so."},"freshness":{"pick":{"score":{"asOf":"2026-09-14","ageDays":43,"stale":true,"staleAfterDays":30,"reason":null},"lastReviewed":{"asOf":"2026-09-14","ageDays":43,"stale":true,"staleAfterDays":30,"reason":null},"sourceCheck":{"checkedAt":"2026-10-27","sourceUrls":3,"sourcesReachable":3,"pricingRows":10,"pricingRowsReconfirmed":8,"note":"A source re-check confirms the page is reachable and, for pricing rows, that the stored amount and SKU name appear on it. It does not re-derive or change the score.","reason":null}},"scoreAsOf":"2026-09-14","editorialPriorAsOf":"2026-09-14","pricingLastVerifiedAt":"2026-10-27","pickLastReviewed":"2026-09-14","runnersUpScoreAsOf":["2026-09-14","2026-09-14","2026-09-14","2026-09-14","2026-10-19","2026-09-14"],"contractAsOf":"2026-10-03","pricingShape":"sourced_rows","normalizedUsdPerMillionTokens":null,"normalizationStatus":"ambiguous","normalizationReason":"input_output_split_not_a_single_rate"},"pricingContract":{"normalizedUsdPerMillionTokens":null,"normalizationStatus":"ambiguous","normalizationReason":"input_output_split_not_a_single_rate","shape":"sourced_rows","filterable":false,"priceSortEligible":true,"priceSortKeyUsdPerMillionInputTokens":0.15,"rows":[{"id":"together-llama-33-70b-input","providerId":"together","sku":"Llama 3.3 70B","unit":"usd_per_million_input_tokens","amountUsd":1.04,"currency":"USD","sourceUrl":"https://www.together.ai/pricing","asOf":"2026-10-03","normalizationStatus":"sourced_token_price","normalizedUsdPerMillionTokens":1.04,"normalizationReason":null,"priceSortEligible":true,"qualifier":"Serverless price-per-1M-tokens table. A separate fine-tune table also lists Llama 3.3 70B at different numbers and was not used."},{"id":"together-llama-33-70b-output","providerId":"together","sku":"Llama 3.3 70B","unit":"usd_per_million_output_tokens","amountUsd":1.04,"currency":"USD","sourceUrl":"https://www.together.ai/pricing","asOf":"2026-10-03","normalizationStatus":"sourced_token_price","normalizedUsdPerMillionTokens":1.04,"normalizationReason":"output_token_price_not_a_single_blended_rate","priceSortEligible":false,"qualifier":null},{"id":"together-deepseek-v4-pro-0813-input","providerId":"together","sku":"DeepSeek V4 Pro 0813","unit":"usd_per_million_input_tokens","amountUsd":1.32,"currency":"USD","sourceUrl":"https://www.together.ai/pricing","asOf":"2026-10-03","normalizationStatus":"sourced_token_price","normalizedUsdPerMillionTokens":1.32,"normalizationReason":null,"priceSortEligible":true,"qualifier":"Serverless price-per-1M-tokens table. A separate fine-tune table also lists Llama 3.3 70B at different numbers and was not used."},{"id":"together-deepseek-v4-pro-0813-output","providerId":"together","sku":"DeepSeek V4 Pro 0813","unit":"usd_per_million_output_tokens","amountUsd":3.96,"currency":"USD","sourceUrl":"https://www.together.ai/pricing","asOf":"2026-10-03","normalizationStatus":"sourced_token_price","normalizedUsdPerMillionTokens":3.96,"normalizationReason":"output_token_price_not_a_single_blended_rate","priceSortEligible":false,"qualifier":null},{"id":"together-deepseek-v4-pro-0813-cached","providerId":"together","sku":"DeepSeek V4 Pro 0813","unit":"usd_per_million_cached_input_tokens","amountUsd":0.13,"currency":"USD","sourceUrl":"https://www.together.ai/pricing","asOf":"2026-10-03","normalizationStatus":"sourced_token_price","normalizedUsdPerMillionTokens":0.13,"normalizationReason":"cached_input_not_a_single_rate","priceSortEligible":false,"qualifier":null},{"id":"together-gpt-oss-120b-input","providerId":"together","sku":"gpt-oss-120B","unit":"usd_per_million_input_tokens","amountUsd":0.15,"currency":"USD","sourceUrl":"https://www.together.ai/pricing","asOf":"2026-10-03","normalizationStatus":"sourced_token_price","normalizedUsdPerMillionTokens":0.15,"normalizationReason":null,"priceSortEligible":true,"qualifier":"Serverless price-per-1M-tokens table. A separate fine-tune table also lists Llama 3.3 70B at different numbers and was not used."},{"id":"together-gpt-oss-120b-output","providerId":"together","sku":"gpt-oss-120B","unit":"usd_per_million_output_tokens","amountUsd":0.6,"currency":"USD","sourceUrl":"https://www.together.ai/pricing","asOf":"2026-10-03","normalizationStatus":"sourced_token_price","normalizedUsdPerMillionTokens":0.6,"normalizationReason":"output_token_price_not_a_single_blended_rate","priceSortEligible":false,"qualifier":null},{"id":"together-glm-53-flash-input","providerId":"together","sku":"GLM-5.3-Flash","unit":"usd_per_million_input_tokens","amountUsd":0.15,"currency":"USD","sourceUrl":"https://www.together.ai/pricing","asOf":"2026-10-03","normalizationStatus":"sourced_token_price","normalizedUsdPerMillionTokens":0.15,"normalizationReason":null,"priceSortEligible":true,"qualifier":"Serverless price-per-1M-tokens table. A separate fine-tune table also lists Llama 3.3 70B at different numbers and was not used."},{"id":"together-glm-53-flash-output","providerId":"together","sku":"GLM-5.3-Flash","unit":"usd_per_million_output_tokens","amountUsd":0.5,"currency":"USD","sourceUrl":"https://www.together.ai/pricing","asOf":"2026-10-03","normalizationStatus":"sourced_token_price","normalizedUsdPerMillionTokens":0.5,"normalizationReason":"output_token_price_not_a_single_blended_rate","priceSortEligible":false,"qualifier":null},{"id":"together-glm-53-flash-cached","providerId":"together","sku":"GLM-5.3-Flash","unit":"usd_per_million_cached_input_tokens","amountUsd":0.03,"currency":"USD","sourceUrl":"https://www.together.ai/pricing","asOf":"2026-10-03","normalizationStatus":"sourced_token_price","normalizedUsdPerMillionTokens":0.03,"normalizationReason":"cached_input_not_a_single_rate","priceSortEligible":false,"qualifier":null}],"notesExcerpt":"As of 2026-09-14: Open-model inference + fine-tune on one OpenAI-compatible API; competitive per-token rates (often ~$0.15–$0.90/MTok class depending on model). Strong for Llama/Qwen/DeepSeek families; Dedicated endpoints available.","notesCheck":{"status":"no_known_contradiction","pairs":[],"review":null,"note":null},"lastReviewed":"2026-09-14","scoreAsOf":"2026-09-14","website":"https://www.together.ai","doNotParse":"Do not parse pricingNotes. Use rows only. normalizedUsdPerMillionTokens is a single number only when this provider has exactly one sourced token row. Input/output, context tiers, and non-token units stay null at this level.","gap":"Partial coverage. Providers without a fetched SKU are omitted. See GET /api/pricing."},"links":{"jtbd":"/jtbds/finetune","jtbdApi":"/api/jtbds/finetune","leaderboard":"/api/leaderboards?jtbd=finetune","methodology":"/api/methodology","badge":"/api/badge?type=depot-pick&jtbd=finetune","decide":"/api/decide?jtbd=finetune","humanPage":"/jtbds/finetune","alternatives":"/api/alternatives?provider=together","canonical":"/api/canonical","quote":"/api/quote?jtbd=finetune","falsify":"/api/falsify?jtbd=finetune"},"constraintBundle":{"task":null,"jtbd":"finetune","quotaExhausted":[],"degraded":[],"healthy":[],"budgetUsd":null,"maxLatencyMs":null,"region":null,"privacy":null,"exit":[],"existing":[]},"stackContext":{"host":null,"framework":null,"repo":null,"lang":null,"existing":[],"volume":null,"sender":null,"detectedFromTask":[]},"keep":null,"contextAdvice":null,"communityFeedback":{"pickOnThisJob":{"failed":0,"degraded":0,"worked":0,"insights":0,"withVerifiedReceipt":0,"failureModes":{},"latestAt":null,"note":"Agent-reported claims (unverified unless a receipt is verified by Depot's re-check). They never change scores or picks."},"outcomes":{"job":{"n":0,"worked":0,"degraded":0,"failed":0,"failureModes":{},"verifiedReports":0,"contributors":0,"excluded":{"seed_editorial":0,"test":0,"notFeedback":0},"reason":"no_submitted_outcome_reports_yet"},"pick":{"providerId":"together","n":0,"worked":0,"degraded":0,"failed":0,"failureModes":{},"evidenceTiers":{"agent_claim":0,"self_declared":0,"reference_checked":0,"outcome_verified":0},"reason":"no_submitted_outcome_reports_yet"},"scoresMutated":false,"more":"/api/outcomes?jtbd=finetune"},"report":"POST /api/feedback {agentIdentity, outcome: failed|degraded|worked, detail, providerId, jtbdId, failureModeId?, receipt?}"}}