DeepSeek V4 Flash Sets ARC-AGI Cost-Performance Record
ARC Prize Foundation verifies the model's benchmark scores at low per-task cost.
ARC Prize Foundation posts state DeepSeek V4 Flash scores 89.0% on ARC-AGI-1 at $0.02 per task and 61.4% on ARC-AGI-2 at $0.04 per task. Together AI calls the model the current cost-performance frontier. Baseten notes its role as inference partner and describes the model as strong on intelligence and cost. Multiple replies on X repeat the verified figures and place the results on a new Pareto frontier.
DeepSeek V4 Flash from @deepseek_ai on ARC-AGI (Verified): - ARC-AGI-2: 61.4%, $0.04/task - ARC-AGI-1: 89.0%, $0.02/task DeepSeek V4 Flash sets the new standard on the cost-to-performance Pareto frontier.
Reactions
17 postsTA 
Together AI@togethercomputeDeepSeek V4 Flash 0728 really is the current cost-performance frontier!
- likes: 122
- replies: 6
- bookmarks: 23
- reposts: 3
GK Greg Kamradt@GregKamradtOutstanding cost-to-performance from DeepSeek GPT-5.6 Luna (Max) performance for a 1/4th of the cost on ARC-AGI
- likes: 139
- replies: 12
- bookmarks: 20
- reposts: 10
GK Greg Kamradt@GregKamradtWhy does the curve zigzag? I asked the same question
- likes: 6
- replies: 1
- bookmarks: 3
- reposts: 0
LA Lisan al Gaib@scaling01I also already tried V4 Flash it is a good model and it's basically free
- likes: 236
- replies: 6
- bookmarks: 14
- reposts: 7
AP ARC Prize@arcprizeDeepSeek V4 Flash from @deepseek_ai on ARC-AGI (Verified): - ARC-AGI-2: 61.4%, $0.04/task - ARC-AGI-1: 89.0%, $0.02/task DeepSeek V4 Flash sets the new standard on the cost-to-performance Pareto frontier.
- likes: 1.8K
- replies: 64
- bookmarks: 312
- reposts: 145
BA Baseten@basetenWe're proud to be @arcprize's inference partner as they test the impact of models like DeepSeek V4 Flash and Kimi K3. Also: more evidence that DeepSeek V4 Flash is a strong model, in terms of both intelligence and cost. Try it out:
- likes: 31
- replies: 0
- bookmarks: 4
- reposts: 3
