Best Fine Tuning Companies

Vendor profileChecked Nov 8, 2026

Fireworks AI

Fireworks AI, Inc. · fireworks.ai · Last checked:

  • Managed fine-tuning API
  • Training API
  • Inference deployment

Company status

Active, privately held.

Sources 1, 2, 3, 4, 5 · checked 2026-11-08

Capabilities

  • Managed SFT and preference tuning (DPO/ORPO) with LoRA (rank up to 128) or full-parameter training
  • Training API: write training loops in Python with custom losses; serverless trainer supports LoRA SFT, DPO, ORPO, RL, and distillation
  • Managed reinforcement fine-tuning (RFT) jobs have been removed; RL now runs on the Training API
  • LoRA adapters can be downloaded with firectl model download
  • Trained models serve on dedicated deployments (live merge or multi-LoRA)

Sources 1, 2, 3, 4, 5 · checked 2026-11-08

Who it tends to fit

  • Teams fine-tuning open models from 1B to over 300B parameters with published per-token rates
  • Teams that need HIPAA (with BAA) or ISO 42001 evidence

Sources 1, 2, 3, 4, 5 · checked 2026-11-08

Pricing

Managed training per 1M training tokens: up to 16B parameters $0.50 LoRA SFT, $1.00 LoRA DPO, $1.00 full SFT, $2.00 full DPO; 16.1B-80B $3.00 / $6.00 / $6.00 / $12.00; 80B-300B $6.00 / $12.00 / $12.00 / $24.00; over 300B $10.00 / $20.00 / $20.00 / $40.00. Dedicated Training API per GPU-hour: H100 and H200 $8.00, B200 $13.00, B300 $15.00, GB300 $20.00.

Sources 1, 2, 3, 4, 5 · checked 2026-11-08

Published certifications

  • SOC 2 Type II
  • ISO 27001
  • ISO 27701
  • ISO 42001
  • HIPAA (with BAA)

Listed as stated by the vendor. Scope can differ by product or subsidiary; request current reports.

Sources 1, 2, 3, 4, 5 · checked 2026-11-08

Limitations to check

  • Trained models serve on dedicated deployments; serverless per-token serving of your own adapter is not available
  • Managed RFT no longer accepts jobs; RL requires the Training API
  • Region-restricted dedicated training costs 1.5x

Sources 1, 2, 3, 4, 5 · checked 2026-11-08

Sources

  1. Training prices per 1M tokens by model size (LoRA/full, SFT/DPO), GPU-hour rates; SOC 2 Type II, ISO 27001, 27701, 42001, HIPAA with BAAfireworks.ai/pricingChecked 2026-11-08
  2. SFT and DPO/ORPO; LoRA rank max 128; rank 0 requests full-parameter training; dedicated deployments only for trained modelsdocs.fireworks.ai/fine-tuning/fine-tuning-modelsChecked 2026-11-08
  3. Managed RFT removed; RL runs on the Training APIdocs.fireworks.ai/fine-tuning/reinforcement-fine-tuning-modelsChecked 2026-11-08
  4. Training API: Python training loops, custom losses, LoRA SFT, DPO, ORPO, RL, distillationdocs.fireworks.ai/fine-tuning/training-api/introductionChecked 2026-11-08
  5. LoRA adapters downloadable with firectl model download; live merge and multi-LoRA deploymentdocs.fireworks.ai/fine-tuning/deploying-lorasChecked 2026-11-08

Report a correction: [email protected]. Facts are re-checked at least every 90 days.