BenchLeader

GPT-5.6 Terra vs Muse Spark 1.2

Verdict
  • GPT-5.6 Terra (xhigh) leads on quality: 64.2 vs 62.6.
  • GPT-5.6 Terra (xhigh) is stronger in agents & tools, human preference, instruction following, maths, multimodal.
  • Muse Spark 1.2 is stronger in coding, composite, knowledge, reasoning.
  • Muse Spark 1.2 is 2.3× cheaper ($2.00 vs $4.50 per 1M blended).
  • Muse Spark 1.2 streams 2.4× faster (219 vs 93 tokens per second).
MetricGPT-5.6 Terra (xhigh)Muse Spark 1.2
BenchLeader Index64.262.6
Agents & tools score69.6
Coding score63.766.5
Composite score77.179.1
Human preference score66.3
Instruction following score65.0
Knowledge score60.976.9
Long context score66.066.0
Maths score71.054.3
Multimodal score63.0
Reasoning score58.270.2
Blended price $/M$4.50$2.00
Output speed93 tok/s219 tok/s
Time to first answer32.5 s23.7 s
Context window1M1.0M
SimpleBench48.9%74.5%
SciCode51.6%
ProofBench74.0%43.0%
Epoch Capabilities Index155.5
LMArena Text1466
LMArena Hard Prompts1491
LMArena Coding1519
LMArena WebDev1521
LMArena Vision1269
LMArena Agent1.5
AA Intelligence Index38.239.8
IFBench66.3%
AA-LCR79.0%79.0%
MMMU-Pro79.5%
AA-Omniscience-3.027.2
Terminal-Bench Hard62.9%
GPQA Diamond (AA)90.8%90.4%
Humanity's Last Exam (AA)41.9%45.5%
SciCode (AA)52.3%57.4%
τ²-Bench Telecom (AA)80.4%
LiveCodeBench85.9%
MMLU-Pro86.7%
IOI65.3%49.5%
CorpFin65.3%
TaxEval76.2%
GPQA Diamond (Vals)90.9%
ARC-AGI-194.0%
ARC-AGI-274.2%
ARC-AGI-30.7%

Data as of 2026-09-09. Best configuration of each model; every score links to its source on the model pages.