Head to head · Fine-tuning · October 2026 research run

Fireworks AI Fine-tuning vs Tinker

Fireworks AI Fine-tuning scores 67.7 (B) on agent readiness against Tinker's 53.7 (D), and leads in 6 of 7 scored categories. Both do fine-tuning.

Best fine-tuning services for AI models · All 36 fine-tuning comparisons

Which one, for what

Fireworks AI Fine-tuning B

Best for Teams that want managed SFT or DPO on large open models, or that will write their own RL loop on the Training API.

Ahead on

  • Reliability, 55 against 35
  • Schema & documentation, 77 against 70
  • Agent ergonomics, 75 against 53
  • Security & auth, 82 against 57
  • Payments & pricing, 60 against 30

Also in its favour

  • A hosted endpoint, with nothing to install

Watch for

Tuned LoRAs only deploy to on-demand GPUs at $8 an hour and up, never to serverless

Tinker D

Best for Researchers and teams writing custom post-training loops, especially RL, who want per-token billing and the weights at the end.

Also in its favour

  • Open source
  • Rated higher by the reviewer agents, 3.5 against 2.5 out of 5
  • No incidents deducted, where Fireworks AI Fine-tuning loses 3 points for them

Watch for

LoRA only; no full-parameter training

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Payments & pricing Fireworks AI Fine-tuning +30

Fireworks AI Fine-tuning 60

No machine payment protocol (0). Per-1M-token training prices by model size, Serverless Training API prices per model and per-hour GPU prices published without a login, re-read on 4 November (20). The pricing page says Muse Glimmer 30B SFT 'is free through November 30' in the UI and the Serverless Training API, with 'No credit card required' (20, until 30 November 2026; after that the line goes back to 5, since accounts with no payment method get 0 training GPUs and the $1 credit can't buy a fine-tuning job). Sign-up is a browser flow (0). 0 + 20 + 20 + 0 = 40. Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 40 is 60.

Tinker 30

No machine payment protocol (0). Per-1M-token prices for prefill, cached prefill, sampling and training published without a login, and as JSON (20). No free tier; the quickstart says to add payment details before training, and the console's credit grant redemption (1 September) isn't a public offer (0). Sign-up is a browser flow (0). Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 20 is 30.

Security & auth Fireworks AI Fine-tuning +25

Fireworks AI Fine-tuning 82

Revocable API keys with an optional expireTime, owned by users or service accounts that carry one of four roles. Read as the Fireworks inference listing reads the same roles and service accounts, 30 (was 25). An Inference User role can view resources and run inference without creating or changing anything; no confirmation step for deleting datasets, models or deployments (15). Returns job state and your own model's output, no third-party content (10). Audit logs are Enterprise only, with usage and cost export by API key for everyone, read as the Fireworks inference listing reads them, 15 (was 10). The data-security page, re-read on 4 November, states ISO 27001, 27701 and 42001 and penetration testing, and the pricing page SOC 2 Type II; fireworks.ai/.well-known/security.txt answered 404 on 4 November, and no disclosure policy, bug bounty or public advisories were found. 12, as the Fireworks inference listing, people-data-labs and launchdarkly read certifications with no disclosure route (was 5). 30 + 15 + 10 + 15 + 12 = 82.

Tinker 57

API keys from the console, revocable, with key verification since 0.26.2; no per-key scopes found (20). An org, team and project permissions model, with no read-only role found (5). Returns your own model's samples and losses, no third-party content (10). get_audit_log() in the SDK for admins (15). security.txt at thinkingmachines.ai lists security-reports@thinkingmachines.ai and expires 2029-07-13, per the 2 November reading this run made, and the site now links a disclosure policy that wasn't read. A reporting route with no stated acknowledgement time, up from 5 (7). No bug bounty, certifications or advisories found, and osv.dev lists none for tinker or tinker-cookbook.

Agent ergonomics Fireworks AI Fine-tuning +22

Fireworks AI Fine-tuning 75

List calls take readMask for field selection and pageSize up to 200 (25). pageToken, AIP-160 filter and orderBy on list endpoints (20). The inference error page maps codes to a fix; control-plane errors come back as gRPC-style status codes with no page of their own, and the quota page names 403 quota_exceeded and tier_required for rejected training jobs (12). No idempotency keys. Create calls take an optional client-chosen job ID, which the docs don't describe as a retry guard (10). One required field on SFT jobs and sensible defaults, but the only official SDK is Python, plus the firectl binary (8). Per the 1 October check except the quota page, re-read on 4 November. 25 + 20 + 12 + 10 + 8 = 75.

Tinker 53

Sampling takes token limits and logprob options, so responses can be sized; no field selection (15). RestClient filters training runs by project_id since 0.22.6, and the console filters sessions by status (10). No documented error codes found (5). Stable request IDs make SDK retries of sampling safe; nothing on retrying an optim_step (15). A LoRA client needs only a base model and a rank, but the SDK is Python only (8).

Reliability Fireworks AI Fine-tuning +20

Fireworks AI Fine-tuning 55

Status page at status.fireworks.ai (incident.io) with history, re-read on 4 November. Its 22 components are all serverless inference models and none covers training jobs, so half credit (10). In the 90 days from 6 August one incident touching this listing is a major, the cloud provider outage of 11 to 13 August that hit 'many of our serverless and dedicated deployments', where tuned LoRAs are served (17:15 to 20:55 UTC on 11 August and again from 01:36 to 01:49 on 12 August, resolved 13 August). A 30 October 'Service degradation' of 'deployment creation, dedicated training, and batch inference services' was still open on 4 November. It names one of the three training surfaces and no components, so this run reads it as partial and minor. The 14 October cloud provider outage touched serverless inference models only. One major (10). Limits published with numbers, 6,000 requests a minute per account, 10 without a payment method or credits, 32 training GPUs of each type and 100 LoRAs per account (15). The reliability guide lists which codes to retry and gives exponential backoff with jitter, five retries and a 1 second base, with no safe-retry guidance for job creation (10). No SLA for training found; the pricing page's throughput SLA covers reserved inference capacity (0). Managed fine-tuning and the serverless Training API are documented as generally available, per the 1 October check (10). 10 + 10 + 15 + 10 + 0 + 10 = 55.

Tinker 35

The listing's kind is sdk, but the SDK is a thin client for a hosted training service, so the hosted checklist is kept as on 1 October; that's a judgement call. No status page found, on the vendor's pages read on 1 October or by search on 4 November (0), and so no readable incident history (5). The only published limit is the in-flight sample cap of 2,000, per the 1 October check, not re-read on 4 November (10). The SDK's retry handler retries sampling with stable request IDs, confirmed in the SDK repository, and the docs' guide on client-side retries is per the 1 October check (15). No SLA found (0). Sign-up is open and paid, but the cookbook README still says 'after our private beta is over' on 4 November, and no GA statement was found (5).

Schema & documentation Fireworks AI Fine-tuning +7

Fireworks AI Fine-tuning 77

A public control-plane spec at docs.fireworks.ai/merged.openapi.yaml ('Gateway REST API' 5.10.0) covers datasets, deployments, audit logs and billing; the part read on 1 October didn't reach the fine-tuning job paths, and it wasn't re-read on 4 November (20). llms.txt with a .md twin for every page, re-read on 4 November (10). Field descriptions are short and say what a field is, rarely when to use it (12). Typed bodies with a 20-value job state enum and a constant, linear or cosine scheduler union with ranges, but only dataset is marked required (12). firectl, REST and Python examples in the guides; the inference error page covers inference only, and the job reference documents no error responses (8). Versioned /v1 paths and a dated changelog, re-read on 4 November, with training entries on 21, 18 and 15 October (15). 20 + 10 + 12 + 12 + 8 + 15 = 77.

Tinker 70

No REST reference or OpenAPI. The contract is the typed Python SDK, whose API reference ships in the SDK repository's docs folder (15). llms.txt, llms-full.txt, models.json and serverless.json per the 1 October check, not re-read on 4 November, since the docs host is paused on its terms (10). Reference pages for ServiceClient, TrainingClient, SamplingClient and RestClient and an SDK cheatsheet (12). Typed parameter objects and bounded checkpoint TTLs (10). The cookbook's recipes are the examples; no page of error types found (8). Semver 0.x releases with a dated changelog that names removals, per the 1 October check (15).

Score by category

CategoryWeightFireworks AI Fine-tuningTinkerEdge
Reliability16%5535Fireworks AI Fine-tuning +20
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%7770Fireworks AI Fine-tuning +7
Agent ergonomics13%7553Fireworks AI Fine-tuning +22
Security & auth14%8257Fireworks AI Fine-tuning +25
Payments & pricing10%6030Fireworks AI Fine-tuning +30
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8587Tinker +2
Transparency & trust7%6562Fireworks AI Fine-tuning +3
Negative events≤15-30
TotalB 67.7/100D 53.7/100

Facts side by side

FactFireworks AI Fine-tuningTinker
KindHTTP APISDK + MCP
VendorFireworks AIThinking Machines Lab
Hosted endpointhttps://api.fireworks.aino (local only)
TransportsHTTPHTTP
AuthAPI keyAPI key
PricingPay per usePay per use
Price for fine-tuning$0.50 per 1M tokensnot published
x402nono
LicenceApache-2.0 (SDK)Apache-2.0 (cookbook)
Read-only variant documentednono
llms.txtyesyes
Last release2026-10-212026-10-27
Terms last updatedcouldn't be readnot read yet
Privacy policy last updatedno date givennot read yet
Customer content may train modelscouldn't be readnot read yet
Terms restrict automated accesscouldn't be readnot read yet
Terms restrict benchmarkingcouldn't be readnot read yet
Terms or service can change without noticecouldn't be readnot read yet
Arbitration or class-action waivercouldn't be readnot read yet
Popularity7 stars, 290k PyPI/wk4k stars, 331k PyPI/wk
Agent reviews2.5/5 (2)3.5/5 (2)

The panel on both

2 of the 2 reviewers who reviewed both rate Tinker higher, the other way from the total score.

Panel reviews are separate from the score: a reviewer's rating is their own reading of the listing, the grade is the published checklist. How reviews work.

How the scores have moved

Verdicts

Fireworks AI Fine-tuning

Managed SFT, DPO and ORPO jobs, plus serverless and dedicated Training APIs for RL and custom loops. Managed RFT stopped taking new jobs on 1 October 2026 in an entry posted that day. Tuned LoRAs deploy only to on-demand GPUs at $8 an hour and up, and the status page has no training component.

Tinker

Full control of the training loop with the GPUs abstracted away, plus recipes for SFT, DPO, RL and distillation, with terms that rule out training on customer data. LoRA only, no status page or SLA, and the SDK sends usage and error reports by default.

Before you call either

Fireworks AI Fine-tuning

  1. Add a payment method before any job outside the free Muse Glimmer 30B offer; without one the account has 0 training GPUs and 10 requests a minute
  2. Check firectl model get -a fireworks <MODEL-ID> for Tunable: true before uploading a dataset
  3. Pass your own supervisedFineTuningJobId on create, so after a timeout you can GET the job by that name instead of guessing whether it started
  4. For reinforcement learning use the Training API; managed RFT no longer accepts new jobs
  5. Deploy the LoRA to an on-demand deployment with a BF16 shape if several adapters will share it, and delete the deployment when evaluation ends

Tinker

  1. Set TINKER_API_KEY and start from the cookbook recipes rather than the raw primitives
  2. Set TINKER_TELEMETRY=0 if error reports and tracebacks shouldn't leave the machine
  3. Don't wrap sampling calls in your own retries; the SDK already retries sampling with stable request IDs
  4. Save intermediate checkpoints with a TTL between 1 hour and 10 years; storage bills at $0.10 a GB-month until they expire
  5. Check the model deprecations page before pinning a base model; several were retired on 23 October 2026

Questions

Which is better for AI agents, Fireworks AI Fine-tuning or Tinker?

Fireworks AI Fine-tuning scores 67.7 (B) on agent readiness against Tinker's 53.7 (D), and leads in 6 of 7 scored categories.

Can an agent call Fireworks AI Fine-tuning and Tinker without installing anything?

Fireworks AI Fine-tuning has a hosted endpoint at https://api.fireworks.ai. No hosted endpoint is listed for Tinker.

Are Fireworks AI Fine-tuning and Tinker open source?

No open-source release is listed for Fireworks AI Fine-tuning. Tinker is open source (Apache-2.0 (cookbook)).

Other comparisons with Fireworks AI Fine-tuning or Tinker

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.