Gemma 4 26B A4B vs GLM-4.7-Flash
Wins 4 of 6 areas
Reasoning · Facts · Long documents · Following instructions
Wins 1 of 6 areas
Agents
Gemma 4 26B A4B is the stronger all-rounder.GLM-4.7-Flash is cheaper and better at agents.
Scores updated · 24 tests both models report · How we compare
Where each one wins
Tests won in each of the six 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.
- FactsGetting facts right instead of making them up30Gemma 4 26B A4B3 of 3 tests
- ReasoningHard problems that need careful thinking20Gemma 4 26B A4B2 of 3 tests · 1 tie
- Long documentsFinding answers in very long texts10Gemma 4 26B A4B1 of 1 test
- Following instructionsDoing exactly what it is asked10Gemma 4 26B A4B1 of 1 test
- AgentsCarrying out multi-step tasks on its own02GLM-4.7-Flash2 of 2 tests
- CodingWriting and fixing software22Even2 each
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.
GLM-4.7-Flash costs 11% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Gemma 4 26B A4B pulls ahead
- Reasons across sets of long documentsAA-LCR+24points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+21.1points ahead
- Recent programming contest problemsLiveCodeBench v6+13.1points ahead
Where GLM-4.7-Flash pulls ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+41.8points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+16.5points ahead
- Real work tasks from 44 professionsGDPVal+13.5points ahead
Every test, side by side
All 24 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- SWE-bench VerifiedGLM-4.7-Flash by 41.817.459.2+41.8
- LiveCodeBench v6Gemma 4 26B A4B by 13.177.164+13.1
- Terminal-Bench HardGLM-4.7-Flash by 8.413.622+8.4
- SciCodeGemma 4 26B A4B by 6.34033.7+6.3
AgentsGLM-4.7-Flash
- BrowseCompGLM-4.7-Flash by 16.526.342.8+16.5
- GDPValGLM-4.7-Flash by 13.53.416.9+13.5
ReasoningGemma 4 26B A4B
- GPQA DiamondGemma 4 26B A4B by 21.179.258.1+21.1
- Humanity's Last ExamGemma 4 26B A4B by 11.719.37.6+11.7
- CritPttie00.3tie
FactsGemma 4 26B A4B
- AA-Omniscience · Non-hallucinationGemma 4 26B A4B by 7.513.66.1+7.5
- Vectara HHEM hallucination ratelower is betterGemma 4 26B A4B by 4.15.29.3+4.1
- AA-Omniscience · AccuracyGemma 4 26B A4B by 2.919.116.1+2.9
Following instructionsGemma 4 26B A4B
- IFBenchGemma 4 26B A4B by 11.672.460.8+11.6
Other results10 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.
- τ²-Bench Telecom (AA run)GLM-4.7-Flash by 55.243.698.8+55.2
- AA Agentic IndexGLM-4.7-Flash by 351146+35
- Artificial Analysis Coding IndexGemma 4 26B A4B by 13.439.325.9+13.4
- AA-OmniscienceGemma 4 26B A4B by 11.8-50.8-62.6+11.8
- τ²-BenchGLM-4.7-Flash by 11.368.279.5+11.3
- vectara_answer_rateGemma 4 26B A4B by 8.299.891.6+8.2
- GPQA (unspecified)Gemma 4 26B A4B by 7.182.375.2+7.1
- vectara_avg_summary_lengthGLM-4.7-Flash by 4.767.171.8+4.7
- vectara_factual_consistencyGemma 4 26B A4B by 4.194.890.7+4.1
- AA IntelligenceGemma 4 26B A4B by 1.816.714.9+1.8
Questions people ask
Which is better, Gemma 4 26B A4B or GLM-4.7-Flash?
Gemma 4 26B A4B wins four of the six areas where both have results: reasoning, facts, long documents and following instructions. GLM-4.7-Flash wins agents, and costs 11% less. They are level on coding.
Which is better for coding?
Neither. They win 2 coding tests each of the 4 both models report.
Which is cheaper?
Gemma 4 26B A4B costs $0.13 per million input tokens and $0.40 per million output tokens; GLM-4.7-Flash costs $0.07 and $0.40. That makes GLM-4.7-Flash about 11% cheaper for the same work.
How do you compare the two?
We use the 24 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 10 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.