DeepSeek-V4-Pro vs Hy3-preview
Wins 4 of 7 areas
Coding · Agents · Reasoning · Following instructions
Wins 1 of 7 areas
Long documents
DeepSeek-V4-Pro wins more areas, narrowly.Hy3-preview is cheaper and better at long documents.
Scores updated · 30 tests both models report · How we compare
Where each one wins
Tests won in each of the seven areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- ReasoningHard problems that need careful thinking20DeepSeek-V4-Pro2 of 3 tests · 1 tie
- CodingWriting and fixing software32DeepSeek-V4-Pro3 of 7 tests · 2 ties
- AgentsCarrying out multi-step tasks on its own32DeepSeek-V4-Pro3 of 6 tests · 1 tie
- Following instructionsDoing exactly what it is asked10DeepSeek-V4-Pro1 of 1 test
- Long documentsFinding answers in very long texts01Hy3-preview1 of 1 test
- FactsGetting facts right instead of making them up11Even1 each
- MathCompetition and research-level math00Even0 each · 1 tie
Math, long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
Hy3-preview costs 47% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where DeepSeek-V4-Pro pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+28.5points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+14.4points ahead
- Complex command-line tasks across many fieldsTerminal-Bench 4.0+14.1points ahead
Where Hy3-preview pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+20points ahead
- Multi-step tasks using many real software toolsMCP Atlas+5.5points ahead
- Reasons across sets of long documentsAA-LCR+4.3points ahead
Every test, side by side
All 30 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V4-Pro
- Terminal-Bench HardDeepSeek-V4-Pro by 14.446.231.8+14.4
- LMArena · WebDevHy3-preview by 44 rating points14641508+44 rating
- SWE-bench VerifiedDeepSeek-V4-Pro by 2.680.678+2.6
- SWE-bench ProHy3-preview by 2.555.457.9+2.5
- SciCodeDeepSeek-V4-Pro by 2.250.848.6+2.2
- SWE-bench Multilingualtie76.275.8tie
- Terminal-Bench 2.1tie6464.4tie
AgentsDeepSeek-V4-Pro
- Terminal-Bench 4.0DeepSeek-V4-Pro by 14.114.60.5+14.1
- τ-Bench V3 · BankingDeepSeek-V4-Pro by 7.230.122.9+7.2
- MCP AtlasHy3-preview by 5.573.679.1+5.5
- GDPValDeepSeek-V4-Pro by 4.632.928.3+4.6
- AA ApexAgentsHy3-preview by 1.324.325.6+1.3
- BrowseComptie83.484.2tie
ReasoningDeepSeek-V4-Pro
- CritPtDeepSeek-V4-Pro by 812.94.9+8
- Humanity's Last ExamDeepSeek-V4-Pro by 437.533.5+4
- GPQA Diamondtie88.889.7tie
FactsEven
- AA-Omniscience · Non-hallucinationHy3-preview by 205.925.9+20
- AA-Omniscience · AccuracyDeepSeek-V4-Pro by 114332+11
Following instructionsDeepSeek-V4-Pro
- IFBenchDeepSeek-V4-Pro by 28.576.548+28.5
Other results9 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- MathArena Apex (Pass@1)DeepSeek-V4-Pro by 51.590.238.7+51.5
- τ²-Bench Telecom (AA run)DeepSeek-V4-Pro by 28.796.267.5+28.7
- DeepSWEHy3-preview by 15.212.828+15.2
- AA-OmniscienceDeepSeek-V4-Pro by 7.8-10.7-18.5+7.8
- NL2RepoHy3-preview by 7.138.545.6+7.1
- AA IntelligenceDeepSeek-V4-Pro by 5.130.425.3+5.1
- ToolathlonDeepSeek-V4-Pro by 3.351.848.5+3.3
- AA Agentic IndexDeepSeek-V4-Pro by 2.127.725.6+2.1
- Artificial Analysis Coding Indextie59.458.8tie
Questions people ask
Which is better, DeepSeek-V4-Pro or Hy3-preview?
DeepSeek-V4-Pro wins four of the seven areas where both have results: coding, agents, reasoning and following instructions. Hy3-preview wins long documents, and costs 47% less. They are level on facts and math.
Which is better for coding?
DeepSeek-V4-Pro. It wins 3 of the 7 coding tests both models report; Hy3-preview wins 2, and 2 are ties.
Which is cheaper?
DeepSeek-V4-Pro costs $0.43 per million input tokens and $0.87 per million output tokens; Hy3-preview costs $0.14 and $0.56. That makes Hy3-preview about 47% cheaper for the same work.
How do you compare the two?
We use the 30 benchmark tests both models have published scores on. The verdict counts the 21 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 9 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.