SambaNova reverses its gemma-4-31B-it price cut
SambaNova Cloud reversed its July 2026 cut to gemma-4-31B-it, raising the rate back to $0.38 input / $1.15 output per 1M tokens from $0.22/$0.59, so gpt-oss-120b is now the sole cheapest model on the rate card.
gemma-4-31B-it $0.22 in / $0.59 out per 1M tokens (July 2026 cut, tied with gpt-oss-120b as cheapest on the card)
gemma-4-31B-it $0.38 in / $1.15 out per 1M tokens (reverted to its pre-cut June 2026 rate; gpt-oss-120b is now the sole cheapest model at $0.22/$0.59)
Less than three weeks after cutting gemma-4-31B-it to $0.22 input / $0.59 output per 1M tokens, SambaNova Cloud’s public rate card has reverted the model to $0.38/$1.15 — its pre-cut June 2026 price. The move is a ~73% increase on input and a ~95% increase on output for the same SKU. No other model on the six-model card moved: DeepSeek-V3.1/V3.2 stay at $3.00/$4.50, Meta-Llama-3.3-70B-Instruct at $0.60/$1.20, and MiniMax-M2.7 keeps its $0.06/1M cached-input rate. gpt-oss-120b is now the sole cheapest model on the card at $0.22/$0.59, a distinction it briefly shared with gemma-4-31B-it in July.
August 2026 rate card: gemma-4-31B-it reverts from $0.22/$0.59 back to $0.38/$1.15 input/output, undoing the July 2026 cut after roughly three weeks. All five other models on the card (MiniMax-M2.7, DeepSeek-V3.1, DeepSeek-V3.2, gpt-oss-120b, Meta-Llama-3.3-70B) are unchanged; gpt-oss-120b is now the sole model at the $0.22 price floor.