Model Support

Long-Term Support (LTS)

Tune Studio guarantees support for at least six months with one-month advance notification before any changes.

Model NameSource URL
tune-blobTune Assistant (not public)
Meta-Llama-3-8B-Instructhttps://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct
Meta-Llama-3-70B-Instructhttps://huggingface.co/meta-llama/Meta-Llama-3-70B-Instruct
mixtral-8x7b-inst-v0–1–32khttps://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1

Experimental Support

Support not guaranteed long-term but one-week advance notification provided before any changes.

Model NameSource URL
llama–3–lumimaid–70bhttps://huggingface.co/NeverSleep/Llama-3-Lumimaid-70B-v0.1
Llama–3–70B-Instruct-Nitrohttps://openrouter.ai/models/meta-llama/llama-3-70b-instruct:nitro
tune-wizardlm–2–8x22bhttps://huggingface.co/dreamgen/WizardLM-2-8x22B
tune-mythomax-l2–13bhttps://huggingface.co/Gryphe/MythoMax-L2-13b
neversleep-llama-3-lumimaid-8bhttps://huggingface.co/NeverSleep/Llama-3-Lumimaid-8B-v0.1
openrouter-goliath-120b-4khttps://huggingface.co/alpindale/goliath-120b
gemma-7b-ithttps://huggingface.co/google/gemma-7b-it
openhermes-2-5-m7b-8khttps://huggingface.co/teknium/OpenHermes-2.5-Mistral-7B

Hardware Pricing

We Support a couple of generally available hardware and some that are offered through support.

This is applicable to:

  1. Models deployed from huggingface
  2. Models deploy from bucket
  3. Finetune jobs
Hardware NamePrice per Hour (USD)Availibility
NVIDIA L41.4Public
NVIDIA A100as per contractSupport
NVIDIA H100as per contractSupport

Public Model Pricing

This pricing is applicable to the models that are available publicly on Tune Studio for API usage.

Model NamePrice per Million Tokens (USD)
gpt-4o15
openrouter-goliath-120b-4k12
tune-mythomax-l2-13b1.125
neversleep-llama-3-lumimaid-8b1.125
llama-3-lumimaid-70b4.5
Llama-3-70B-Instruct-Nitro1.2
Meta-Llama-3-70B-Instruct1.2
phi-3-medium-128k-instruct1
mixtral-8x7b-inst-v0-1-32k0.5
fast-cohereaya 8b0.3
gemma-7b-it0.3
Meta-Llama-3-8B-Instruct0.3
openhermes-2-5-m7b-8k0.3
mistral-7b-instruct-v0-30.07

We do not charge per token for models deployed on GPUs or for models integrated by you.