GLM-4.7 vs Step 3.5 Flash
Wins 4 of 7 areas
Coding · Reasoning · Long documents · Following instructions
Wins 2 of 7 areas
Agents · Math
GLM-4.7 wins more areas, narrowly.Step 3.5 Flash is cheaper and better at agents.
Scores updated · 31 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 thinking20GLM-4.72 of 3 tests · 1 tie
- CodingWriting and fixing software21GLM-4.72 of 4 tests · 1 tie
- Long documentsFinding answers in very long texts10GLM-4.71 of 1 test
- Following instructionsDoing exactly what it is asked10GLM-4.71 of 1 test
- AgentsCarrying out multi-step tasks on its own02Step 3.5 Flash2 of 2 tests
- MathCompetition and research-level math01Step 3.5 Flash1 of 1 test
- FactsGetting facts right instead of making them up11Even1 each
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.
Step 3.5 Flash costs 86% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where GLM-4.7 pulls ahead
- Reasons across sets of long documentsAA-LCR+21points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+6.3points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+5.7points ahead
Where Step 3.5 Flash pulls ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+17points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+7.3points ahead
- Olympiad math problems with short answersIMOAnswerBench+3.4points ahead
Every test, side by side
All 31 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGLM-4.7
- SciCodeGLM-4.7 by 4.745.140.4+4.7
- Terminal-Bench HardGLM-4.7 by 4.531.827.3+4.5
- LiveCodeBench v6Step 3.5 Flash by 1.584.986.4+1.5
- SWE-bench Verifiedtie73.874.4tie
AgentsStep 3.5 Flash
- BrowseCompStep 3.5 Flash by 175269+17
- GDPValStep 3.5 Flash by 1.825.827.6+1.8
ReasoningGLM-4.7
- Humanity's Last ExamGLM-4.7 by 6.327.421.1+6.3
- GPQA DiamondGLM-4.7 by 2.885.983.1+2.8
- CritPttie1.72.5tie
FactsEven
- AA-Omniscience · Non-hallucinationStep 3.5 Flash by 7.3714.3+7.3
- AA-Omniscience · AccuracyGLM-4.7 by 5.729.323.6+5.7
Other results17 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.
- AA Agentic IndexStep 3.5 Flash by 25.826.252+25.8
- GAIA (no file)Step 3.5 Flash by 22.661.984.5+22.6
- Artificial Analysis Coding IndexGLM-4.7 by 13.745.331.6+13.7
- Terminal-Bench 2.0Step 3.5 Flash by 104151+10
- AA IntelligenceGLM-4.7 by 5.622.216.6+5.6
- AA-OmniscienceGLM-4.7 by 5.4-36.4-41.8+5.4
- xbench-DeepSearchStep 3.5 Flash by 452.356.3+4
- xbench-DeepSearch (2025.10)Step 3.5 Flash by 452.356.3+4
- ResearchRubricsStep 3.5 Flash by 3.36265.3+3.3
- AIME 2025Step 3.5 Flash by 1.695.797.3+1.6
- browsecomp_with_context_managerStep 3.5 Flash by 1.567.569+1.5
- τ²-Bench Telecom (AA run)GLM-4.7 by 1.595.994.4+1.5
- HMMT Feb. 2025Step 3.5 Flash by 1.397.198.4+1.3
- τ²-Benchtie87.488.2tie
- HMMT Nov. 2025tie93.594tie
- BrowseComp-ZHtie66.666.9tie
- MMLU-Protie84.384.4tie
Questions people ask
Which is better, GLM-4.7 or Step 3.5 Flash?
GLM-4.7 wins four of the seven areas where both have results: coding, reasoning, long documents and following instructions. Step 3.5 Flash wins agents and math, and costs 86% less. They are level on facts.
Which is better for coding?
GLM-4.7. It wins 2 of the 4 coding tests both models report; Step 3.5 Flash wins 1, and 1 is a tie.
Which is cheaper?
GLM-4.7 costs $0.60 per million input tokens and $2.20 per million output tokens; Step 3.5 Flash costs $0.10 and $0.30. That makes Step 3.5 Flash about 86% cheaper for the same work.
How do you compare the two?
We use the 31 benchmark tests both models have published scores on. The verdict counts the 14 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 17 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.