SPACEXAI
4/5 workers samplingPOSSIBLE_MODEL_CHANGE · gpt-5
inference

Together

San Francisco · us-east, us-west, europe · Open-weight hosting. H100/H200 fleet.

Reported availability
99.72%
Shown separately from GBU samples
Live endpoints
3
Regions covered
3
ModelAPI idTTFTtok/sEWR$/correctGBU availEvidence
Llama 4 405B
Together
meta-llama/Llama-4-405B-Instruct280 ms1102.97$0.03299.72%LIVE_MEASURED
Llama 4 70B
Together
meta-llama/Llama-4-70B-Instruct140 ms19532.3$0.006499.75%LIVE_MEASURED
Qwen3-Max
Together
Qwen/Qwen3-Max260 ms922.56$0.03699.70%LIVE_MEASURED