Skip to content

Fine-tuningThinking Machines Lab

Tinker

Managed fine-tuning API with low-level control over the training loop.

Category
Fine-tuning
Pricing
PAID
Source
Proprietary
Hosting
Cloud
Platforms
API
Models
Multi-model
Verified
Jun 10, 2026

Tinker is Thinking Machines Lab's training API for fine-tuning open-weight LLMs. It exposes low-level primitives — forward_backward, optim_step, sample — so researchers keep full control of data and algorithms while the service handles distributed GPU scheduling and failure recovery. LoRA-based runs cover models from small Llamas up to large mixture-of-experts like Qwen-235B and Kimi K2, and trained weights can be downloaded.

Capabilities 2

What it actually does — grouped by capability family.

  • Fine-tuning / training (primary capability)
  • Multi-model access (secondary capability)

Pros & cons

  • Exposes forward_backward, optim_step, sample
  • Large MoE models supported
  • Downloadable trained weights
  • No GPU infra to manage
  • LoRA-based only (no full fine-tune)
  • Usage-based, no free tier
  • Python SDK only

Tags

Further reading

Alternatives to Tinker (4)

View all Fine-tuning →
  • View OpenPipe details
    Fine-tuningFREEMIUM

    OpenPipe

    OpenPipe

    Replace frontier-model spend with a fine-tuned small model.

    Captures your production OpenAI / Anthropic calls, builds a dataset, fine-tunes a small open-weights model on your traffic, then serves the swap behind your existing SDK. The pitch: 10x cost reduction at parity.

    Uses your production logs as training data
    Needs enough quality traffic to distill
    • fine-tuning
    • cost-reduction
    • drop-in
    • open-weights
    Open
  • View Predibase details
    Fine-tuningPAID

    Predibase

    Predibase (Rubrik)

    Fine-tune open-source LLMs and serve them in production.

    Predibase is an enterprise platform for fine-tuning open-source models and serving them in production. It pairs a post-training stack — supervised fine-tuning plus an end-to-end reinforcement fine-tuning (RFT) flow — with an optimized inference engine, and its open-source LoRAX framework serves many fine-tuned LoRA adapters from a single GPU. Runs as managed SaaS or inside your own VPC.

    Fine-tune + serve in one place
    Enterprise-priced
    • fine-tuning
    • lora
    • rft
    • inference
    • +1
    Open
  • View Unsloth details
    Fine-tuningFREEMIUMOpen core

    Unsloth

    Unsloth AI

    Fine-tune open LLMs faster with far less VRAM.

    An open-source (Apache-2.0) framework for fine-tuning and running open-weight models with custom CUDA kernels — roughly 2x faster training and large VRAM savings, so 7B–13B models fit on a single consumer GPU. Free tier runs on Colab/Kaggle or locally; Pro and Enterprise tiers add multi-GPU and multi-node speedups. Exports to GGUF/Safetensors for llama.cpp, vLLM, and Ollama.

    LoRA, QLoRA, and full fine-tuning
    Multi-GPU speedups are paid tiers
    • fine-tuning
    • lora
    • open-source
    • training
    Open
  • View Axolotl details
    Fine-tuningFREEOSS

    Axolotl

    Axolotl AI

    Open-source post-training for LLMs — LoRA to RL, all from one YAML config.

    An open-source (Apache-2.0) framework that streamlines post-training for open-weight models: full fine-tuning, LoRA/QLoRA, preference tuning (DPO, IPO, KTO, ORPO), reinforcement learning (GRPO), reward modeling and quantization-aware training, configured through a single YAML file with no scripting. Wraps Hugging Face Transformers, PEFT, TRL and DeepSpeed, and supports dozens of model families including multimodal vision and audio models.

    Apache-2.0 with 12k+ GitHub stars
    Needs your own GPUs or cloud compute
    • fine-tuning
    • lora
    • rlhf
    • open-source
    Open
Open Tinker