GLM 5.3 Flash run record

GLM 5.3 Flash

z-ai/glm-5.3-flash

starter@v1HostedOpenRouterserved by StreamLakeFP81M context2026-10-03 18:45 UTC

Overall score

48 of 153 test cases fully passed

Throughput

63.2tok/s

Mean per test case, 60 tok/s over the run

Output per test

2.8Ktokens

Answer plus reasoning

Reasoning per test

2.5Ktokens

89% of the output

Total tokens

1.3M

882.2K in, 427.9K out

Run time

2h 21m

55.5s per caseRequest time, summed

Cost

$0.19

$0.0013 per caseAt the provider's list price

Every test case

153 test cases in 8 tasks

One square per test case. Select a square to read the answer.

PassedPartial creditFailed

10Agents2 of 21 passed

2 of 21 passed

48Structured Extraction10 of 21 passed

10 of 21 passed

54Coding6 of 21 passed

6 of 21 passed

37Grounded Answers8 of 21 passed

8 of 21 passed

43Instruction Following6 of 21 passed

6 of 21 passed

33Business Logic7 of 21 passed

7 of 21 passed

55Security & permissions9 of 21 passed

9 of 21 passed

0Diagnosis & investigation0 of 6 passed

0 of 6 passed

Speed and output per test case

One bar per test case. Hover to inspect a case, select it to read the answer.

ThroughputPeak 99 tok/s

Output tokensReasoningAnswer

Test by test

Every answer with its reasoning and the checks it had to pass.

Reproduce this run

The prompts stay private so models can't train on them. The hashes pin the exact suite and config this run used.

Serve the modelOpenRouterHosted

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"z-ai/glm-5.3-flash","provider":{"order":["streamlake/fp8"],"allow_fallbacks":false},"messages":[{"role":"user","content":"Hello"}]}'

The request the benchmark sent, pinned to the provider that served it.

Record id
Suite hash, starter@v1
Config hash

Configuration

Where the suite ran, and the token budget each task got.

Deployment: Hosted
API: OpenRouter
Serving provider: StreamLake
Precision: FP8
Context window: 1M
Temperature: 0
Bench version: v1.0.0
Ran at: 2026-10-03 18:45 UTC

Provider parameters

{
  "provider": {
    "order": [
      "streamlake/fp8"
    ],
    "allow_fallbacks": false
  }
}

Token budget per task
Task	Reasoning	Max output per call
Agents	4.1Khigh	12.3K
Structured Extraction	4.1Khigh	12.3K
Coding	4.1Khigh	12.3K
Grounded Answers	4.1Khigh	12.3K
Instruction Following	4.1Khigh	12.3K
Business Logic	4.1Khigh	12.3K
Security & permissions	4.1Khigh	12.3K
Diagnosis & investigation	4.1Khigh	12.3K

GLM 5.3 Flash Compare with another model

Every test case

Speed and output per test case

Test by test

10AgentsRuns multi-step jobs through tools, within a written policy2 of 21 passedp90 latency 1m 1s

48Structured ExtractionTurns logs, diffs and records into exact JSON10 of 21 passedp90 latency 2m 53s

54CodingWrites code and SQL that pass hidden tests6 of 21 passedp90 latency 4m 43s

37Grounded AnswersAnswers from the given documents with sources, never guesses8 of 21 passedp90 latency 3m 41s

43Instruction FollowingApplies every rule, including ones changed mid-conversation6 of 21 passedp90 latency 1m 7s

33Business LogicComputes billing, scheduling and pricing rules exactly7 of 21 passedp90 latency 3m 0s

55Security & permissionsIgnores planted instructions, still does the allowed work9 of 21 passedp90 latency 56.2s

0Diagnosis & investigationChooses diagnostic checks, repairs supported causes and verifies recovery0 of 6 passedp90 latency 36.6s

Reproduce this run

Configuration