Inference
EN

GPT-6 Astra vs Hy4

Change models
GPT-6 AstraOpenAI
Hy4Tencent
IndexAll categories, out of 100.
78.0 ±1.9 best
72.0 ±7.8
GapIndex points below the top.
Top
−6.0 · no clear gap
CoverageIndex weight that has results.
100% best
65%
SpeedOutput tokens per second.
34/s
52/s best
EloLMArena rating and rank.
1440
—
ReleasedMonth of release.
2026-09
2026-11
Input / 1MUS dollars per 1M input tokens.
$10.00
$0.417 best
Output / 1MUS dollars per 1M output tokens.
$50.00
$1.25 best
Batch output / 1MBatch price per 1M output tokens.
$25.00
—
1K chats2K tokens in, 500 out, each.
$45.00
$1.46 best
1K documents20K tokens in, 1K out, each.
$250.00
$9.59 best
1K agent tasks200K tokens in, 20K out, each.
$3,000.00
$108.41 best
ContextMaximum tokens in one request.
1.05M
1.05M
AgenticMulti-step tasks with tools.
74.6 best
74.5
CodingCode writing and repair.
76.9 best
68.4
ReasoningLogic problems and puzzles.
83.8
—
MultimodalTasks with images and text.
71.2 best
57.4
KnowledgeFacts and expert knowledge.
78.5
83.7 best
MultilingualTasks in many languages.
73.6
—
InstructionTasks with strict prompt rules.
79.5
—
MathMath problems.
78.0
—

Benchmarks

14 shared
Only GPT-6 Astra: 124Only Hy4: 7
GPT-6 Astra
Hy4
Vibe Code Bench v1.1Coding
89.6% best8 / 123
81.0%30 / 123
Legal Research BenchLegal
39.4%37 / 90
41.3% best30 / 90
Tax Agent BenchFinance
20.7%52 / 82
34.4% best17 / 82
Terminal-Bench 4.0Agentic
59.6% best5 / 61
39.4%15 / 61
DeepSWEAgentic
74.1%11 / 59
74.9% best8 / 59
Agent Arena steerabilityAgentic
14.8 best5 / 54
8.814 / 54
Agent Arena task outcomeAgentic
14.0 best6 / 54
6.712 / 54
Humanity's Last Exam with toolsAgentic
57.2%13 / 36
59.1% best10 / 36
HLAB-Vals-1.1Legal
7.1%25 / 35
12.5% best17 / 35
AutomationBenchAgentic
41.4%18 / 32
56.2% best4 / 32
FrontierCode 2.0Coding
42.8% best3 / 27
31.0%9 / 27
Agents' Last ExamAgentic
59.3% best2 / 21
41.0%8 / 21
OSWorld 2.1 (all 108 tasks)Agentic
75.8%7 / 11
78.9% best5 / 11
Terminal-Bench-Science 0.1Agentic
64.6% best3 / 11
19.8%11 / 11