Qwen3.8 27B (high) leads on average score with 8.7 vs 8.6. Claude Haiku 5.5 (low) has the lower benchmark cost at $0.061 vs ~$0.298. Claude Haiku 5.5 (low) is faster at 7.51s vs 166.34s, with pass rates of 76.8% vs 78.3%.
Last updated at:
2026-10-07
Compared models
Rank
#64
Total Output Tokens
61,650
Response Time (avg)
7.51s
Total Cost
$0.061
Candidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.
Rank
#56
Total Output Tokens
677,080
Response Time (avg)
166.34s
Total Cost
~$0.298Estimated GPU electricity cost only. It excludes the rest of the computer and hardware ownership. NVIDIA GeForce RTX 3090: 3.0102 h × 0.300 kW × €0.2896/kWh × 1.138 USD/EUR = ~$0.297613.
Recommended modelClaude Haiku 5.5 (low)
Its score stays close to the best score here (8.6 vs 8.7), while costing about 4.9x less than Qwen3.8 27B (high).
Qwen3.8 27BQwen3.8 27BhighCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.Release: 2026-08-14Free Available
Qwen3.8 27BQwen3.8 27BhighCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.Release: 2026-08-14Free Available
Score
8.6Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
8.7Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
Rank
#64
#56
Reliability
10.0First-attempt success score: 10.0 means no retryable target API or rate-limit failures before successful calls; tracked failures lower the score.…
9.7First-attempt success score: 10.0 means no retryable target API or rate-limit failures before successful calls; tracked failures lower the score.…
Consistency
8.9Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
9.2Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
A test is fully passed only if every run passed for that test.Did not follow instructions: 2No answer: 2Wrong answer: 2Response Time (avg)166.34sResponse Time (max)640.85sResponse Time (total)3825.81sA test is fully passed only if every run passed for that test.…
Attempt pass rate
76.8%Attempt pass rate = passed attempts / total attempts across runs.…
78.3%Attempt pass rate = passed attempts / total attempts across runs.…
Flaky tests
3Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
2Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
Total Runs
69Total Runs…
69Total Runs…
Cost per result
0.378
~1.751Estimated GPU electricity cost only. It excludes the rest of the computer and hardware ownership. NVIDIA GeForce RTX 3090: 3.0102 h × 0.300 kW × €0.2896/kWh × 1.138 USD/EUR = ~$0.297613.
Total Cost
$0.061
~$0.298Estimated GPU electricity cost only. It excludes the rest of the computer and hardware ownership. NVIDIA GeForce RTX 3090: 3.0102 h × 0.300 kW × €0.2896/kWh × 1.138 USD/EUR = ~$0.297613.
Input Price
$0.100 / 1MInput Price…
N/AInput Price…
Output Price
$0.500 / 1MOutput Price…
N/AOutput Price…
Cache Read Price
$0.010 / 1MCache Read Price…
N/ACache Read Price…
Cache Write Price
$0.125 / 1MCache Write Price…
N/ACache Write Price…
Total Input Tokens
296,301Total Input Tokens…
348,967Total Input Tokens…
Output Tokens
10,672Output Tokens…
1,005Output Tokens…
Reasoning Tokens
50,978Reasoning Tokens…
676,075Reasoning Tokens…
Response Time (avg)
7.51sResponse Time (avg)…
166.34sResponse Time (avg)…
Response Time (max)
44.43sResponse Time (max)…
640.85sResponse Time (max)…
Response Time (total)
172.74sResponse Time (total)…
3825.81sResponse Time (total)…
Parameters
~50B
27.3B
Availability
Closed
Weights available
Cache prices apply to input tokens. Reads reuse cached prompts; writes store them and can cost extra. Output tokens use the output price.
Model generation showcase
Hamster playing table tennis
Prompt: Create a detailed SVG illustration of a hamster playing table tennis.
#64 Claude Haiku 5.5
low
Cost
$0.001
Time
7.4s
Tokens
1,851 tok
#56 Qwen3.8 27B
highCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.
Cost
~$0.008Estimated GPU electricity cost only. It excludes the rest of the computer and hardware ownership. NVIDIA GeForce RTX 3090: 0.0750 h × 0.300 kW × €0.2896/kWh × 1.138 USD/EUR = ~$0.007419.
10.0Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
100.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No failed answers.Response Time (avg)44.43sResponse Time (max)44.43sResponse Time (total)44.43sA test is fully passed only if every run passed for that test.…
44.43sResponse Time (avg)…
153,966Total Input Tokens…
2,991Output Tokens…
3,811Reasoning Tokens…
Qwen3.8 27BCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.
10.0Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
100.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No failed answers.Response Time (avg)132.27sResponse Time (max)132.27sResponse Time (total)132.27sA test is fully passed only if every run passed for that test.…
8.7Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
7.9Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
91.7%Attempt pass rate = passed attempts / total attempts across runs.…
1Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.Wrong answer: 1Response Time (avg)2.16sResponse Time (max)3.03sResponse Time (total)8.62sA test is fully passed only if every run passed for that test.…
2.16sResponse Time (avg)…
858Total Input Tokens…
483Output Tokens…
1,480Reasoning Tokens…
Qwen3.8 27BCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.
10.0Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
100.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No failed answers.Response Time (avg)8.10sResponse Time (max)21.48sResponse Time (total)32.40sA test is fully passed only if every run passed for that test.…
7.9Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
7.5Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
77.8%Attempt pass rate = passed attempts / total attempts across runs.…
1Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.Did not follow instructions: 1Response Time (avg)7.30sResponse Time (max)8.92sResponse Time (total)21.89sA test is fully passed only if every run passed for that test.…
7.30sResponse Time (avg)…
10,608Total Input Tokens…
531Output Tokens…
11,618Reasoning Tokens…
Qwen3.8 27BCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.
8.1Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
7.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
88.9%Attempt pass rate = passed attempts / total attempts across runs.…
1Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.Did not follow instructions: 1Response Time (avg)248.62sResponse Time (max)458.25sResponse Time (total)745.85sA test is fully passed only if every run passed for that test.…
10.0Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
100.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No failed answers.Response Time (avg)23.94sResponse Time (max)27.39sResponse Time (total)47.88sA test is fully passed only if every run passed for that test.…
23.94sResponse Time (avg)…
104,796Total Input Tokens…
4,741Output Tokens…
19,543Reasoning Tokens…
Qwen3.8 27BCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.
10.0Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
100.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No failed answers.Response Time (avg)185.57sResponse Time (max)333.21sResponse Time (total)371.15sA test is fully passed only if every run passed for that test.…
10.0Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
100.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No failed answers.Response Time (avg)3.45sResponse Time (max)3.70sResponse Time (total)6.89sA test is fully passed only if every run passed for that test.…
3.45sResponse Time (avg)…
10,515Total Input Tokens…
312Output Tokens…
1,609Reasoning Tokens…
Qwen3.8 27BCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.
10.0Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
100.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No failed answers.Response Time (avg)34.85sResponse Time (max)57.71sResponse Time (total)69.71sA test is fully passed only if every run passed for that test.…
5.9Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
7.2Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
55.6%Attempt pass rate = passed attempts / total attempts across runs.…
1Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.Extra formatting: 1Wrong answer: 1Response Time (avg)7.35sResponse Time (max)7.77sResponse Time (total)22.06sA test is fully passed only if every run passed for that test.…
7.35sResponse Time (avg)…
990Total Input Tokens…
64Output Tokens…
9,207Reasoning Tokens…
Qwen3.8 27BCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.
5.3Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
33.3%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No answer: 1Wrong answer: 1Response Time (avg)526.74sResponse Time (max)640.85sResponse Time (total)1580.21sA test is fully passed only if every run passed for that test.…
4.6Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
9.8Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
0.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.Wrong answer: 1Response Time (avg)1.38sResponse Time (max)1.38sResponse Time (total)1.38sA test is fully passed only if every run passed for that test.…
1.38sResponse Time (avg)…
714Total Input Tokens…
319Output Tokens…
0Reasoning Tokens…
Qwen3.8 27BCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.
5.0Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
0.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.Did not follow instructions: 1Response Time (avg)8.45sResponse Time (max)8.45sResponse Time (total)8.45sA test is fully passed only if every run passed for that test.…
9.8Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
100.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No failed answers.Response Time (avg)2.20sResponse Time (max)3.24sResponse Time (total)4.40sA test is fully passed only if every run passed for that test.…
2.20sResponse Time (avg)…
921Total Input Tokens…
94Output Tokens…
515Reasoning Tokens…
Qwen3.8 27BCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.
9.8Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
100.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No failed answers.Response Time (avg)8.24sResponse Time (max)10.18sResponse Time (total)16.49sA test is fully passed only if every run passed for that test.…
7.7Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
66.7%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.Did not follow instructions: 1Response Time (avg)2.76sResponse Time (max)3.28sResponse Time (total)8.28sA test is fully passed only if every run passed for that test.…
2.76sResponse Time (avg)…
912Total Input Tokens…
611Output Tokens…
1,783Reasoning Tokens…
Qwen3.8 27BCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.
7.6Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
7.2Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
77.8%Attempt pass rate = passed attempts / total attempts across runs.…
1Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No answer: 1Response Time (avg)199.90sResponse Time (max)569.33sResponse Time (total)599.69sA test is fully passed only if every run passed for that test.…
10.0Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
100.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No failed answers.Response Time (avg)4.19sResponse Time (max)4.19sResponse Time (total)4.19sA test is fully passed only if every run passed for that test.…
4.19sResponse Time (avg)…
11,757Total Input Tokens…
354Output Tokens…
459Reasoning Tokens…
Qwen3.8 27BCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.
10.0Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
100.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No failed answers.Response Time (avg)5.62sResponse Time (max)5.62sResponse Time (total)5.62sA test is fully passed only if every run passed for that test.…
3.0Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
0.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.No answer: 1Response Time (avg)2.70sResponse Time (max)2.70sResponse Time (total)2.70sA test is fully passed only if every run passed for that test.…
2.70sResponse Time (avg)…
264Total Input Tokens…
172Output Tokens…
953Reasoning Tokens…
Qwen3.8 27BCandidate responses were generated on local hardware. OpenRouter was used only for judging. qwen3.8:latest · Q4_K_M · Ollama 0.32.12 · NVIDIA GeForce RTX 3090.
3.0Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
10.0Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
0.0%Attempt pass rate = passed attempts / total attempts across runs.…
0Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
A test is fully passed only if every run passed for that test.Wrong answer: 1Response Time (avg)263.98sResponse Time (max)263.98sResponse Time (total)263.98sA test is fully passed only if every run passed for that test.…