Across 10 comparable benchmarks, GLM-5.3 has better reported results on 10 and worse on 0.
Moderate evidence10 of 10 shared results countedHow we compare
Price basis: International rates vs listed rates. The price section below shows the applicable periods.
| Fact | GLM-5.3 | Qwen3.6-27B |
|---|---|---|
| Context window | 1M tokens | 128K tokens |
| Input / 1M | $1.40 | $0.600 |
| Output / 1M | $4.40 | $3.60 |
| Blended (4:1) | $1.18 | $1.20 |
| Status | Available | Available |
| Released | August 14, 2026 | April 22, 2026 |
Blended is 4:1 input to output with 90% cache hits at listed rates β set your own mix.
Mean of 4 shared benchmarks grouped by capability on a 0β100 scale. How averages are built
GLM-5.3 ahead on 1 of 1 counted
GLM-5.3 ahead on 1 of 1 counted
GLM-5.3 ahead on 2 of 2 counted
No fixed-scale average β see the rows in Benchmark results
GLM-5.3 ahead on 1 of 1 counted
GLM-5.3 ahead on 1 of 1 counted
GLM-5.3 ahead on 1 of 1 counted
No fixed-scale average β see the rows in Benchmark results
No shared comparable results for Abstract reasoning, Computer & tool use, Research & search, Cybersecurity.
Published scores, with matching modes where available. Gaps show GLM-5.3 relative to Qwen3.6-27B. Reading the results
| Benchmark | GLM-5.3 | Qwen3.6-27B | Gap |
|---|---|---|---|
| Vibe Code Bench v1.1Code Generation & Editing Β· Pass rate (%) | 78.1Max Β· tools | 11.9β Unspecified Reasoning Β· tools | +66.2 pts#37 vs #100 of 121 |
| FrontierMath v2Mathematics Β· Accuracy (verification_code) | 68.8Max Β· tools | 35.1β Reported best | +33.7 pts#32 vs #66 of 85 |
| CritPtScientific Reasoning Β· Score | 19.1Max Β· no tools | 1.1β Thinking Β· no tools | 17.4Γ#32 vs #103 of 134 |
| SciCodeScientific Computing Β· Score | 59.0Max Β· no tools | 42.8β Thinking Β· no tools | 1.38Γ#15 vs #70 of 98 |
| Context ArenaMemory & Persistence Β· Accuracy (8 needles, 4K-128K context) | 88.5Thinking High Β· no tools | 82.2β Thinking Β· no tools | +6.4 pts#10 vs #18 of 61 |
| GDP.pdfDocuments & Charts Β· Task pass rate (%) | 11.2Max Β· no tools | 11.0β Thinking Β· no tools | +0.2 pts#55 vs #58 of 96 |
| ΟΒ³-BankingService Workflows Β· Score | 50.3Max Β· tools | 16.7β Thinking Β· tools | 3.01Γ#4 vs #67 of 120 |
| HLEKnowledge Exams Β· Accuracy | 62.5Max Β· tools | 24.0β Thinking Β· no tools | +38.5 pts#9 vs #94 of 151 |
| ECICapability Indices Β· ECI score (capability index, higher is better) | 156Reported best | 146Reported best | 1.06Γ#24 vs #67 of 168 |
| LiveBenchCross-capability Suites Β· Accuracy | 76.1Reported best | 64.0β Normal Β· no tools | +12.1 pts#28 vs #72 of 106 |
Code Generation & Editing Β· Pass rate (%)
Mathematics Β· Accuracy (verification_code)
Memory & Persistence Β· Accuracy (8 needles, 4K-128K context)
Documents & Charts Β· Task pass rate (%)
Service Workflows Β· Score
A dash means no result is available.
Time-of-day providers use peak-hour rates; fixed-price providers use standard rates. Off-peak discounts remain in the published details.
Drag the handle above the bar for cache hits, the one below it for the input : output mix Β· 5% steps
Blended price = effective input rate at 90% cache hits Γ 80% + output rate Γ 20%
GLM-5.3 costs1.7%lessthan Qwen3.6-27B per million tokens.Using international rates vs listed rates at this token mix.
GLM-5.3: International pricing. Z.ai international price (docs.z.ai/guides/overview/pricing) verified 2026-10-22 via browser
Qwen3.6-27B: International pricing. context_length <= 256000 Alibaba Model Studio International (Singapore)
| Price / 1M tokens | GLM-5.3 | Qwen3.6-27B | Ratio |
|---|---|---|---|
| Input | $1.40 | $0.60 | 2.33Γ |
| Output | $4.40 | $3.60 | 1.22Γ |
| Cached input | $0.26 | β | β |
| Cache write | β | β | β |
| Blended | $1.18 | $1.20 | 0.98Γ |
Estimates use the selected token mix. Cache writes, tool fees and long-context surcharges are excluded.
Reported cost per task, paired with the score from the same run. Settings and sources may differ; costs do not affect the verdict.
| Benchmark | GLM-5.3 | Qwen3.6-27B |
|---|---|---|
| AA Coding Agent Index v1.5 | Not reported | |
| AA Intelligence Index v5.0.1 | Not reported | |
| Agent Arena | $0.8526 / task Median Β· Net improvement index (percentage points vs. average model): 0.93 Max (With Tools + Internet) | Not reported |
| Agent Arena Index | $0.9588 / task Median Β· Agent Arena Index rating (index points): 1095 Max (With Tools + Internet) | Not reported |
| CursorBench 4.0 | $5.05 / task Average Β· Task score: 42.6 Max (With Tools) Β· cursorbench-4-0-cursor-agent $3.24 / task Average Β· Task score: 38.0 Thinking High (With Tools) Β· cursorbench-4-0-cursor-agent $2.04 / task Average Β· Task score: 33.3 Thinking Low (With Tools) Β· cursorbench-4-0-cursor-agent | Not reported |
| CWE-bench v0 | Not reported | |
| DeepSWE | Not reported | |
| FrontierCode 2.0 | $5.181 / task Average Β· Weighted rubric score: 16.8 Thinking High (With Tools + Internet) $3.189 / task Average Β· Weighted rubric score: 14.4 Thinking Low (With Tools + Internet) | Not reported |
| FrontierCS 2 | Not reported | |
| FrontierSWE v2 | Not reported | |
| GDPx | Not reported | |
| LiveBench 2.0 | $1.263 / task Average Β· Overall score (equal-weight mean of 7 suite scores, 0-100): 60.7 Max (task-specific tools) | Not reported |
| SAGE | Not reported | |
| Terminal-Bench 4.0 | $9.1247 / task Average Β· Resolution rate (%): 37.4 Max (With Tools) Β· vals-effort-study $8.5536 / task Average Β· Resolution rate (%): 43.4 Thinking High (With Tools) Β· vals-effort-study 3 more settings$5.8894 / task Average Β· Resolution rate (%): 33.3 Thinking Low (With Tools) Β· vals-effort-study $9.37 / task Average Β· Resolution rate (%): 38.9 Unspecified Reasoning (With Tools) Β· vals-mini-swe | Not reported |
| Terminal-Bench 4.1 | Not reported | |
| Vals Index v2.1 | Not reported | |
| Vals Index v2.2 | Not reported | |
| Vibe Code Bench 1-100 | Not reported | |
| Vibe Code Bench v1.1 | Not reported |
| Attribute | GLM-5.3 | Qwen3.6-27B |
|---|---|---|
| Availability | Available | Available |
| Provider | Zhipu AI | Alibaba |
| Model type | Reasoning model | Reasoning model |
| First release | August 14, 2026 | April 22, 2026 |
| Context window | 1M tokens | 128K tokens |
| Max output | 128K tokens | 16K tokens |
| Total parameters | 744B | 27B |
| Active parameters | 40B | 27B |
| Architecture | Mixture of Experts (MoE) | Dense |
| Documented capabilities | Reasoning, Multilingual, Coding, Function calling, Tool use, Long context | Reasoning, Multilingual, Coding, Tool use, Function calling |
| Input modalities | Text | Text, Image, Video |
| Output modalities | Text | Text |
| Weights | Open weights | Open weights |
| Licence | GLM-5.3 License | Qwen License |
| Commercial use | Commercial use allowed, with conditions | Commercial use allowed |
These are reported-score comparisons, not controlled head-to-head tests.
What people found comparing these models, or using either one. Comments also appear on the models' own pages.
Used GLM-5.3 or Qwen3.6-27B? Sign in with Google to share what worked, what did not, and which API or setup you tested.
No community comments on either model yet.
Alternatives with the most benchmark overlap for this matchup.