Claude Haiku 5.5 (medium) vs Inkling 2 Mini (local Q4) (low)
Claude Haiku 5.5 (medium) leads on average score with 7.7 vs 7.6. Claude Haiku 5.5 (medium) has the lower benchmark cost at $0.063 vs ~$0.292. Claude Haiku 5.5 (medium) is faster at 7.71s vs 153.99s, with pass rates of 71.0% vs 66.7%.
Last updated at:
2026-11-22
Compared models
Rank
#197
Total Output Tokens
66,174
Response Time (avg)
7.71s
Total Cost
$0.063
Candidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · NVIDIA GeForce RTX 3090.
Rank
#202
Total Output Tokens
129,600
Response Time (avg)
153.99s
Total Cost
~$0.292Estimated GPU electricity cost only. It excludes the rest of the computer and hardware ownership. NVIDIA GeForce RTX 3090: 2.9515 h × 0.300 kW × €0.2896/kWh × 1.138 USD/EUR = ~$0.291818.
Recommended modelClaude Haiku 5.5 (medium)
It has the best score here (7.7), while costing about 4.6x less than Inkling 2 Mini (local Q4) (low).
Inkling 2 Mini (local Q4)Inkling 2 Mini (local Q4)lowCandidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · NVIDIA GeForce RTX 3090.Release: 2026-11-20
Inkling 2 Mini (local Q4)Inkling 2 Mini (local Q4)lowCandidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · NVIDIA GeForce RTX 3090.Release: 2026-11-20
Score
7.7Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
7.6Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
Rank
#197
#202
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
7.5Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
9.6Consistency 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.Wrong answer: 5Did not follow instructions: 1No answer: 1Timed out: 1Response Time (avg)153.99sResponse Time (max)1963.67sResponse Time (total)3541.85sA test is fully passed only if every run passed for that test.…
Attempt pass rate
71.0%Attempt pass rate = passed attempts / total attempts across runs.…
66.7%Attempt pass rate = passed attempts / total attempts across runs.…
Flaky tests
7Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
1Flaky tests had mixed outcomes across runs (at least one pass and one fail).…
Total Runs
69Total Runs…
69Total Runs…
Cost per result
0.485
~1.946Estimated GPU electricity cost only. It excludes the rest of the computer and hardware ownership. NVIDIA GeForce RTX 3090: 2.9515 h × 0.300 kW × €0.2896/kWh × 1.138 USD/EUR = ~$0.291818.
Total Cost
$0.063
~$0.292Estimated GPU electricity cost only. It excludes the rest of the computer and hardware ownership. NVIDIA GeForce RTX 3090: 2.9515 h × 0.300 kW × €0.2896/kWh × 1.138 USD/EUR = ~$0.291818.
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
298,334Total Input Tokens…
312,366Total Input Tokens…
Output Tokens
10,328Output Tokens…
1,631Output Tokens…
Reasoning Tokens
55,846Reasoning Tokens…
127,969Reasoning Tokens…
Response Time (avg)
7.71sResponse Time (avg)…
153.99sResponse Time (avg)…
Response Time (max)
47.70sResponse Time (max)…
1963.67sResponse Time (max)…
Response Time (total)
177.29sResponse Time (total)…
3541.85sResponse Time (total)…
Parameters
~50B
26.4B total (3.4B active)
Availability
Closed
Open source
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.
#197 Claude Haiku 5.5
medium
Cost
$0.002
Time
17.2s
Tokens
3,524 tok
#202 Thinking Machines: Inkling 2 Mini (local Q4)
lowCandidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · NVIDIA GeForce RTX 3090.
Cost
~$0.003Estimated GPU electricity cost only. It excludes the rest of the computer and hardware ownership. NVIDIA GeForce RTX 3090: 0.0237 h × 0.300 kW × €0.2896/kWh × 1.138 USD/EUR = ~$0.002347.
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)47.70sResponse Time (max)47.70sResponse Time (total)47.70sA test is fully passed only if every run passed for that test.…
47.70sResponse Time (avg)…
163,704Total Input Tokens…
2,705Output Tokens…
6,272Reasoning Tokens…
Inkling 2 Mini (local Q4)Candidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · 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)58.97sResponse Time (max)58.97sResponse Time (total)58.97sA 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)2.36sResponse Time (max)3.39sResponse Time (total)9.44sA test is fully passed only if every run passed for that test.…
2.36sResponse Time (avg)…
858Total Input Tokens…
471Output Tokens…
1,871Reasoning Tokens…
Inkling 2 Mini (local Q4)Candidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · 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)5.40sResponse Time (max)8.07sResponse Time (total)21.60sA test is fully passed only if every run passed for that test.…
6.3Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
4.8Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
66.7%Attempt pass rate = passed attempts / total attempts across runs.…
2Flaky 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: 2Response Time (avg)7.16sResponse Time (max)8.94sResponse Time (total)21.47sA test is fully passed only if every run passed for that test.…
7.16sResponse Time (avg)…
10,608Total Input Tokens…
518Output Tokens…
11,415Reasoning Tokens…
Inkling 2 Mini (local Q4)Candidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · NVIDIA GeForce RTX 3090.
5.3Summarizes 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).…
44.4%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.Timed out: 1Wrong answer: 1Response Time (avg)831.97sResponse Time (max)1963.67sResponse Time (total)2495.90sA test is fully passed only if every run passed for that test.…
7.3Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
5.8Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
83.3%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)20.29sResponse Time (max)26.07sResponse Time (total)40.58sA test is fully passed only if every run passed for that test.…
20.29sResponse Time (avg)…
97,091Total Input Tokens…
4,812Output Tokens…
17,244Reasoning Tokens…
Inkling 2 Mini (local Q4)Candidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · NVIDIA GeForce RTX 3090.
6.5Summarizes 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).…
50.0%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)123.84sResponse Time (max)227.77sResponse Time (total)247.69sA 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)3.43sResponse Time (max)3.56sResponse Time (total)6.87sA test is fully passed only if every run passed for that test.…
3.43sResponse Time (avg)…
10,515Total Input Tokens…
312Output Tokens…
1,666Reasoning Tokens…
Inkling 2 Mini (local Q4)Candidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · 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)20.83sResponse Time (max)21.70sResponse Time (total)41.66sA test is fully passed only if every run passed for that test.…
2.9Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
4.4Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
22.2%Attempt pass rate = passed attempts / total attempts across runs.…
2Flaky 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: 2Wrong answer: 1Response Time (avg)7.96sResponse Time (max)9.19sResponse Time (total)23.89sA test is fully passed only if every run passed for that test.…
7.96sResponse Time (avg)…
990Total Input Tokens…
71Output Tokens…
10,376Reasoning Tokens…
Inkling 2 Mini (local Q4)Candidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · NVIDIA GeForce RTX 3090.
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.Wrong answer: 1Response Time (avg)198.65sResponse Time (max)335.23sResponse Time (total)595.96sA test is fully passed only if every run passed for that test.…
3.8Summarizes broad quality across our full private benchmark suite, so ranking reflects consistent performance.…
2.5Consistency score reflects run-to-run stability (10 = very consistent, even if consistently wrong).…
33.3%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)5.17sResponse Time (max)5.17sResponse Time (total)5.17sA test is fully passed only if every run passed for that test.…
5.17sResponse Time (avg)…
714Total Input Tokens…
192Output Tokens…
1,933Reasoning Tokens…
Inkling 2 Mini (local Q4)Candidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · NVIDIA GeForce RTX 3090.
5.5Summarizes 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)5.55sResponse Time (max)5.55sResponse Time (total)5.55sA 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)3.01sResponse Time (max)3.29sResponse Time (total)6.02sA test is fully passed only if every run passed for that test.…
3.01sResponse Time (avg)…
921Total Input Tokens…
100Output Tokens…
1,278Reasoning Tokens…
Inkling 2 Mini (local Q4)Candidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · 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)5.10sResponse Time (max)5.52sResponse Time (total)10.21sA 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)2.67sResponse Time (max)3.28sResponse Time (total)8.01sA test is fully passed only if every run passed for that test.…
2.67sResponse Time (avg)…
912Total Input Tokens…
568Output Tokens…
2,119Reasoning Tokens…
Inkling 2 Mini (local Q4)Candidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · NVIDIA GeForce RTX 3090.
5.8Summarizes 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).…
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.Did not follow instructions: 1Wrong answer: 1Response Time (avg)10.51sResponse Time (max)14.31sResponse Time (total)31.52sA 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)5.02sResponse Time (max)5.02sResponse Time (total)5.02sA test is fully passed only if every run passed for that test.…
5.02sResponse Time (avg)…
11,757Total Input Tokens…
354Output Tokens…
609Reasoning Tokens…
Inkling 2 Mini (local Q4)Candidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · 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)7.53sResponse Time (max)7.53sResponse Time (total)7.53sA 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)3.12sResponse Time (max)3.12sResponse Time (total)3.12sA test is fully passed only if every run passed for that test.…
3.12sResponse Time (avg)…
264Total Input Tokens…
225Output Tokens…
1,063Reasoning Tokens…
Inkling 2 Mini (local Q4)Candidate responses were generated on local hardware. OpenRouter was used only for judging. inkling-2-mini-ud-q4_k_xl · UD-Q4_K_XL · llama.cpp b12631-e300919cc · 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)25.28sResponse Time (max)25.28sResponse Time (total)25.28sA test is fully passed only if every run passed for that test.…