two model releases shipped on the same day, april 24. it’s worth putting them next to each other.
// openai released GPT-5.5 at $5/$30 per million in/out tokens. closed weights, 1M context, agent stack baked in.
// deepseek released V4 the same day. V4-Pro at $1.74/$3.48, V4-Flash at $0.14/$0.28 per million in/out tokens. 1M context. open weights under MIT license. compressed sparse attention plus heavily compressed attention bringing 1M-token inference to 27% of V3.2’s flops and 10% of its kv cache.
V4-Flash at $0.14 input is 36x cheaper than GPT-5.5. on output, $0.28 vs $30 is over 100x. that’s not “competitive open-source pricing.” that’s a different category.
the closed-source price increase last week made the open-source release the same day feel less coincidental.
// openai released GPT-5.5 at $5/$30 per million in/out tokens. closed weights, 1M context, agent stack baked in.
// deepseek released V4 the same day. V4-Pro at $1.74/$3.48, V4-Flash at $0.14/$0.28 per million in/out tokens. 1M context. open weights under MIT license. compressed sparse attention plus heavily compressed attention bringing 1M-token inference to 27% of V3.2’s flops and 10% of its kv cache.
V4-Flash at $0.14 input is 36x cheaper than GPT-5.5. on output, $0.28 vs $30 is over 100x. that’s not “competitive open-source pricing.” that’s a different category.
the closed-source price increase last week made the open-source release the same day feel less coincidental.