Fabrication bar = red, longer = worse. est = estimated.
Peak throughput and max concurrent = measured using FullBox mk5 + 8xB300.
🇺🇸 GPT-6-Astrathinking:max
OpenAI · USA
99.1Kamiwaza Agent Intelligence score
KAMI frontier99.2%
KAMI medium100.0%
Grounding98.1%
Aggregation97.6%
Fabrication0.0%
Token Efficiency (tokens / correct answer)113,072
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 Gemini-3.8-Flashthinking:high
Google · USA
98.9Kamiwaza Agent Intelligence score
KAMI frontier100.0%
KAMI medium99.4%
Grounding97.6%
Aggregation96.9%
Fabrication0.8%
Token Efficiency (tokens / correct answer)777,076
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 GPT-6-Astrathinking:medium
OpenAI · USA
98.6Kamiwaza Agent Intelligence score
KAMI frontier100.0%
KAMI medium100.0%
Grounding97.2%
Aggregation95.8%
Fabrication2.3%
Token Efficiency (tokens / correct answer)78,561
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 Gemini-3.8-Flashthinking:medium
Google · USA
98.5Kamiwaza Agent Intelligence score
KAMI frontier99.2%
KAMI medium98.8%
Grounding96.9%
Aggregation97.1%
Fabrication0.6%
Token Efficiency (tokens / correct answer)642,867
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 Grok-4.6thinking
xAI · USA
98.4Kamiwaza Agent Intelligence score
KAMI frontier99.2%
KAMI medium99.4%
Grounding95.7%
Aggregation96.1%
Fabrication0.0%
Token Efficiency (tokens / correct answer)148,420
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 Claude-Fable-5.1thinking
Anthropic · USA
97.8Kamiwaza Agent Intelligence score
KAMI frontier96.7%
KAMI medium100.0%
Grounding97.2%
Aggregation95.3%
Fabrication0.7%
Token Efficiency (tokens / correct answer)98,730
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 Muse-Spark-1.3thinking
Meta · USA
97.3Kamiwaza Agent Intelligence score
KAMI frontier95.0%
KAMI medium99.4%
Grounding96.5%
Aggregation96.4%
Fabrication0.1%
Token Efficiency (tokens / correct answer)580,777
Peak throughput (tok/s)—
Max concurrent agents—
🇨🇳 Qwen3.8-Flash-Nextthinking
Alibaba · China
96.3Kamiwaza Agent Intelligence score
KAMI frontier94.2%
KAMI medium99.4%
Grounding95.7%
Aggregation96.3%
Fabrication3.8%
Token Efficiency (tokens / correct answer)993,927
Peak throughput (tok/s)2,660
Max concurrent agents136
🇨🇳 Qwen3.8-2.4T-A95Bthinking:medium
Alibaba · China
96.2Kamiwaza Agent Intelligence score
KAMI frontier94.2%
KAMI medium98.1%
Grounding95.3%
Aggregation95.7%
Fabrication1.3%
Token Efficiency (tokens / correct answer)743,303
Peak throughput (tok/s)—
Max concurrent agents—
🇨🇳 Qwen3.8-2.4T-A95Bthinking:xhigh
Alibaba · China
96.0Kamiwaza Agent Intelligence score
KAMI frontier92.5%
KAMI medium99.4%
Grounding95.6%
Aggregation96.3%
Fabrication2.4%
Token Efficiency (tokens / correct answer)812,777
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 GPT-5.6-solthinking
OpenAI · USA
95.8Kamiwaza Agent Intelligence score
KAMI frontier95.8%
KAMI medium99.4%
Grounding96.6%
Aggregation95.0%
Fabrication10.3%
Token Efficiency (tokens / correct answer)221,148
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 Claude-Opus-5thinking
Anthropic · USA
95.6Kamiwaza Agent Intelligence score
KAMI frontier98.3%
KAMI medium92.5%
Grounding97.0%
Aggregation97.0%
Fabrication7.6%
Token Efficiency (tokens / correct answer) est80,757
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 Claude-Sonnet-5thinking
Anthropic · USA
94.7Kamiwaza Agent Intelligence score
KAMI frontier91.7%
KAMI medium98.1%
Grounding96.9%
Aggregation91.7%
Fabrication3.9%
Token Efficiency (tokens / correct answer) est318,842
Peak throughput (tok/s)—
Max concurrent agents—
🇨🇳 DeepSeek-V4.1-Flashthinking:max
DeepSeek · China
94.5Kamiwaza Agent Intelligence score
KAMI frontier90.8%
KAMI medium95.0%
Grounding94.7%
Aggregation97.0%
Fabrication1.5%
Token Efficiency (tokens / correct answer)4,042,542
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 GPT-5.6-solthinking:max
OpenAI · USA
94.2Kamiwaza Agent Intelligence score
KAMI frontier91.7%
KAMI medium93.8%
Grounding98.6%
Aggregation97.3%
Fabrication7.7%
Token Efficiency (tokens / correct answer)118,292
Peak throughput (tok/s)—
Max concurrent agents—
🇨🇳 DeepSeek-V4-Flash-0731thinking
DeepSeek · China
92.7Kamiwaza Agent Intelligence score
KAMI frontier87.5%
KAMI medium96.2%
Grounding94.4%
Aggregation95.2%
Fabrication7.0%
Token Efficiency (tokens / correct answer)222,881
Peak throughput (tok/s)2,765
Max concurrent agents128
🇨🇳 GLM-5.2thinking
Z.ai (Zhipu) · China
92.4Kamiwaza Agent Intelligence score
KAMI frontier80.8%
KAMI medium98.8%
Grounding96.0%
Aggregation95.4%
Fabrication1.5%
Token Efficiency (tokens / correct answer)166,075
Peak throughput (tok/s)739
Max concurrent agents12
🇨🇳 Kimi-K3thinking
Moonshot AI · China
91.1Kamiwaza Agent Intelligence score
KAMI frontier80.8%
KAMI medium97.5%
Grounding97.3%
Aggregation96.8%
Fabrication10.7%
Token Efficiency (tokens / correct answer)109,873
Peak throughput (tok/s)253
Max concurrent agents24
🇨🇳 Qwen3.8-27Bthinking
Alibaba · China
88.9Kamiwaza Agent Intelligence score
KAMI frontier71.7%
KAMI medium95.6%
Grounding95.1%
Aggregation97.0%
Fabrication1.9%
Token Efficiency (tokens / correct answer)164,117
Peak throughput (tok/s)5,539
Max concurrent agents152
🇨🇳 DeepSeek-V4.1-Flashthinking:high
DeepSeek · China
88.8Kamiwaza Agent Intelligence score
KAMI frontier76.7%
KAMI medium90.0%
Grounding94.7%
Aggregation95.6%
Fabrication1.9%
Token Efficiency (tokens / correct answer)844,260
Peak throughput (tok/s)—
Max concurrent agents—
🇨🇳 DeepSeek-V4-Flash-0731instruct
DeepSeek · China
88.4Kamiwaza Agent Intelligence score
KAMI frontier89.2%
KAMI medium93.8%
Grounding92.6%
Aggregation71.6%
Fabrication9.4%
Token Efficiency (tokens / correct answer)272,201
Peak throughput (tok/s)2,765
Max concurrent agents128
🇺🇸 GPT-5.6-terrathinking
OpenAI · USA
87.2Kamiwaza Agent Intelligence score
KAMI frontier77.5%
KAMI medium88.1%
Grounding96.3%
Aggregation90.7%
Fabrication7.8%
Token Efficiency (tokens / correct answer)219,779
Peak throughput (tok/s)—
Max concurrent agents—
🇨🇳 GLM-5.2instruct
Z.ai (Zhipu) · China
86.6Kamiwaza Agent Intelligence score
KAMI frontier65.8%
KAMI medium98.1%
Grounding96.5%
Aggregation88.8%
Fabrication3.0%
Token Efficiency (tokens / correct answer)139,994
Peak throughput (tok/s)739
Max concurrent agents12
🇨🇳 Qwen3.6-27Binstruct
Alibaba · China
86.6Kamiwaza Agent Intelligence score
KAMI frontier70.0%
KAMI medium96.9%
Grounding96.2%
Aggregation91.6%
Fabrication12.2%
Token Efficiency (tokens / correct answer)180,868
Peak throughput (tok/s)5,573
Max concurrent agents168
🇨🇳 GLM-5.3thinking:max
Z.ai (Zhipu) · China
84.3Kamiwaza Agent Intelligence score
KAMI frontier58.3%
KAMI medium95.6%
Grounding96.4%
Aggregation96.4%
Fabrication6.5%
Token Efficiency (tokens / correct answer)120,782
Peak throughput (tok/s)1,091
Max concurrent agents56
🇨🇳 MiniMax-M3-MXFP8thinking
MiniMax · China
83.5Kamiwaza Agent Intelligence score
KAMI frontier53.3%
KAMI medium96.2%
Grounding96.4%
Aggregation95.4%
Fabrication2.1%
Token Efficiency (tokens / correct answer)182,760
Peak throughput (tok/s)1,177
Max concurrent agents60
🇺🇸 Muse-Glimmer-30Bthinking
Meta · USA
82.3Kamiwaza Agent Intelligence score
KAMI frontier57.5%
KAMI medium88.1%
Grounding93.0%
Aggregation95.7%
Fabrication1.7%
Token Efficiency (tokens / correct answer)697,384
Peak throughput (tok/s)—
Max concurrent agents—
🇨🇳 Qwen3.6-27Bthinking
Alibaba · China
81.4Kamiwaza Agent Intelligence score
KAMI frontier64.2%
KAMI medium89.4%
Grounding98.0%
Aggregation95.0%
Fabrication27.4%
Token Efficiency (tokens / correct answer)161,930
Peak throughput (tok/s)5,573
Max concurrent agents168
🇨🇳 MiniMax-M3thinking
MiniMax · China
80.8Kamiwaza Agent Intelligence score
KAMI frontier48.3%
KAMI medium91.2%
Grounding96.2%
Aggregation95.7%
Fabrication2.1%
Token Efficiency (tokens / correct answer)209,534
Peak throughput (tok/s)1,078
Max concurrent agents44
🇨🇳 MiniMax-M3instruct
MiniMax · China
80.3Kamiwaza Agent Intelligence score
KAMI frontier51.7%
KAMI medium90.0%
Grounding96.2%
Aggregation93.4%
Fabrication7.6%
Token Efficiency (tokens / correct answer)240,305
Peak throughput (tok/s)1,078
Max concurrent agents44
🇨🇳 GLM-5.3-Flashthinking
Z.ai (Zhipu) · China
80.1Kamiwaza Agent Intelligence score
KAMI frontier56.7%
KAMI medium81.9%
Grounding93.2%
Aggregation96.2%
Fabrication5.1%
Token Efficiency (tokens / correct answer)1,085,988
Peak throughput (tok/s)—
Max concurrent agents—
🇨🇳 MiniMax-M3-MXFP8instruct
MiniMax · China
78.9Kamiwaza Agent Intelligence score
KAMI frontier47.5%
KAMI medium90.0%
Grounding96.0%
Aggregation93.5%
Fabrication8.5%
Token Efficiency (tokens / correct answer)212,736
Peak throughput (tok/s)1,177
Max concurrent agents60
🇺🇸 GPT-5.6-lunathinking
OpenAI · USA
78.2Kamiwaza Agent Intelligence score
KAMI frontier49.2%
KAMI medium90.0%
Grounding94.2%
Aggregation87.3%
Fabrication8.3%
Token Efficiency (tokens / correct answer)340,577
Peak throughput (tok/s)—
Max concurrent agents—
🇨🇳 GLM-5.3thinking:low
Z.ai (Zhipu) · China
77.2Kamiwaza Agent Intelligence score
KAMI frontier50.0%
KAMI medium85.0%
Grounding95.3%
Aggregation89.8%
Fabrication12.0%
Token Efficiency (tokens / correct answer)70,907
Peak throughput (tok/s)1,091
Max concurrent agents56
🇨🇳 Qwen3.8-Flash-Nextinstruct
Alibaba · China
73.2Kamiwaza Agent Intelligence score
KAMI frontier36.7%
KAMI medium83.8%
Grounding96.0%
Aggregation94.7%
Fabrication15.6%
Token Efficiency (tokens / correct answer)1,194,605
Peak throughput (tok/s)2,660
Max concurrent agents136
🇺🇸 Nemotron-3-Ultra-550B-A55B-BF16thinking
NVIDIA · USA
72.4Kamiwaza Agent Intelligence score
KAMI frontier39.2%
KAMI medium74.4%
Grounding91.3%
Aggregation90.7%
Fabrication1.4%
Token Efficiency (tokens / correct answer)4,664,083
Peak throughput (tok/s)—
Max concurrent agents—
🇨🇳 Qwen3.8-27Binstruct
Alibaba · China
71.3Kamiwaza Agent Intelligence score
KAMI frontier33.3%
KAMI medium76.9%
Grounding94.9%
Aggregation94.7%
Fabrication8.8%
Token Efficiency (tokens / correct answer)177,243
Peak throughput (tok/s)5,539
Max concurrent agents152
🇺🇸 Inkling-NVFP4thinking
Thinking Machines Lab · USA
70.2Kamiwaza Agent Intelligence score
KAMI frontier35.0%
KAMI medium68.8%
Grounding96.7%
Aggregation92.2%
Fabrication5.6%
Token Efficiency (tokens / correct answer)651,554
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 Gemma-4-31B-itthinking
Google · USA
69.8Kamiwaza Agent Intelligence score
KAMI frontier28.3%
KAMI medium73.8%
Grounding95.0%
Aggregation92.8%
Fabrication2.3%
Token Efficiency (tokens / correct answer)197,292
Peak throughput (tok/s)1,391
Max concurrent agents80
🇫🇷 Mistral-Medium-3.5-128Bthinking
Mistral AI · France
69.1Kamiwaza Agent Intelligence score
KAMI frontier29.2%
KAMI medium80.0%
Grounding89.8%
Aggregation80.9%
Fabrication1.9%
Token Efficiency (tokens / correct answer)439,007
Peak throughput (tok/s)3,114
Max concurrent agents44
🇺🇸 Inklingthinking
Thinking Machines Lab · USA
68.8Kamiwaza Agent Intelligence score
KAMI frontier35.0%
KAMI medium64.4%
Grounding96.4%
Aggregation92.5%
Fabrication7.4%
Token Efficiency (tokens / correct answer)669,328
Peak throughput (tok/s)—
Max concurrent agents—
🇨🇳 Qwen3.6-35B-A3Binstruct
Alibaba · China
68.8Kamiwaza Agent Intelligence score
KAMI frontier30.8%
KAMI medium73.1%
Grounding95.4%
Aggregation90.2%
Fabrication10.5%
Token Efficiency (tokens / correct answer)347,897
Peak throughput (tok/s)12,538
Max concurrent agents320
🇺🇸 Laguna-S-2.1thinking
Poolside AI · USA
68.1Kamiwaza Agent Intelligence score
KAMI frontier55.0%
KAMI medium88.8%
Grounding78.4%
Aggregation46.4%
Fabrication29.0%
Token Efficiency (tokens / correct answer)361,348
Peak throughput (tok/s)10,323
Max concurrent agents400
🇨🇳 Qwen3.5-397B-A17Binstruct
Alibaba · China
67.4Kamiwaza Agent Intelligence score
KAMI frontier20.8%
KAMI medium75.6%
Grounding93.0%
Aggregation93.3%
Fabrication4.4%
Token Efficiency (tokens / correct answer)399,048
Peak throughput (tok/s)1,677
Max concurrent agents112
🇨🇳 Qwen3.5-397B-A17Bthinking
Alibaba · China
66.2Kamiwaza Agent Intelligence score
KAMI frontier12.5%
KAMI medium76.2%
Grounding95.7%
Aggregation93.8%
Fabrication0.4%
Token Efficiency (tokens / correct answer)458,730
Peak throughput (tok/s)1,677
Max concurrent agents112
🇨🇳 Qwen3.6-35B-A3Bthinking
Alibaba · China
66.2Kamiwaza Agent Intelligence score
KAMI frontier29.2%
KAMI medium71.9%
Grounding96.5%
Aggregation92.3%
Fabrication26.0%
Token Efficiency (tokens / correct answer)322,227
Peak throughput (tok/s)12,538
Max concurrent agents320
🇫🇷 Mistral-Medium-3.5-128Binstruct
Mistral AI · France
66.1Kamiwaza Agent Intelligence score
KAMI frontier34.2%
KAMI medium78.8%
Grounding84.8%
Aggregation67.1%
Fabrication10.9%
Token Efficiency (tokens / correct answer)453,135
Peak throughput (tok/s)3,114
Max concurrent agents44
🇨🇳 Qwen3.5-122B-A10Binstruct
Alibaba · China
66.0Kamiwaza Agent Intelligence score
KAMI frontier30.0%
KAMI medium58.8%
Grounding94.0%
Aggregation91.6%
Fabrication3.4%
Token Efficiency (tokens / correct answer)433,821
Peak throughput (tok/s)4,938
Max concurrent agents192
🇨🇳 Qwen3.5-122B-A10Bthinking
Alibaba · China
65.1Kamiwaza Agent Intelligence score
KAMI frontier26.7%
KAMI medium56.9%
Grounding94.7%
Aggregation92.8%
Fabrication1.5%
Token Efficiency (tokens / correct answer)463,557
Peak throughput (tok/s)4,938
Max concurrent agents192
🇺🇸 Inkling-Smallthinking
Thinking Machines Lab · USA
63.6Kamiwaza Agent Intelligence score
KAMI frontier30.0%
KAMI medium57.5%
Grounding92.8%
Aggregation90.9%
Fabrication15.6%
Token Efficiency (tokens / correct answer)276,947
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 Laguna-M.1thinking
Poolside AI · USA
60.6Kamiwaza Agent Intelligence score
KAMI frontier30.8%
KAMI medium69.4%
Grounding82.5%
Aggregation56.1%
Fabrication11.6%
Token Efficiency (tokens / correct answer)791,060
Peak throughput (tok/s)850
Max concurrent agents64
🇨🇳 Qwen3.5-35B-A3Bthinking
Alibaba · China
58.7Kamiwaza Agent Intelligence score
KAMI frontier19.2%
KAMI medium40.6%
Grounding94.4%
Aggregation95.2%
Fabrication4.4%
Token Efficiency (tokens / correct answer)747,719
Peak throughput (tok/s)12,074
Max concurrent agents272
🇨🇳 Qwen3.5-35B-A3Binstruct
Alibaba · China
58.0Kamiwaza Agent Intelligence score
KAMI frontier20.8%
KAMI medium41.2%
Grounding95.0%
Aggregation89.4%
Fabrication8.1%
Token Efficiency (tokens / correct answer)739,632
Peak throughput (tok/s)12,074
Max concurrent agents272
🇺🇸 Gemma-4-31B-itinstruct
Google · USA
55.9Kamiwaza Agent Intelligence score
KAMI frontier5.8%
KAMI medium50.0%
Grounding94.6%
Aggregation86.7%
Fabrication3.7%
Token Efficiency (tokens / correct answer)286,586
Peak throughput (tok/s)1,391
Max concurrent agents80
🇺🇸 Claude-Haiku-4.5thinking
Anthropic · USA
53.8Kamiwaza Agent Intelligence score
KAMI frontier9.2%
KAMI medium43.8%
Grounding93.2%
Aggregation76.6%
Fabrication2.6%
Token Efficiency (tokens / correct answer) est859,350
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 Gemma-4-26B-A4B-itthinking
Google · USA
50.8Kamiwaza Agent Intelligence score
KAMI frontier13.3%
KAMI medium31.2%
Grounding86.5%
Aggregation85.1%
Fabrication11.2%
Token Efficiency (tokens / correct answer)678,713
Peak throughput (tok/s)2,661
Max concurrent agents240
🇺🇸 Gemma-4-26B-A4B-itinstruct
Google · USA
45.5Kamiwaza Agent Intelligence score
KAMI frontier7.5%
KAMI medium23.8%
Grounding82.8%
Aggregation71.8%
Fabrication6.2%
Token Efficiency (tokens / correct answer)910,031
Peak throughput (tok/s)2,661
Max concurrent agents240
🇺🇸 Nemotron-3-Super-120B-A12B-BF16thinking
NVIDIA · USA
45.3Kamiwaza Agent Intelligence score
KAMI frontier7.5%
KAMI medium16.2%
Grounding82.7%
Aggregation82.2%
Fabrication5.1%
Token Efficiency (tokens / correct answer)55,662,220
Peak throughput (tok/s)—
Max concurrent agents—
🇨🇳 Qwen3.5-9Bthinking
Alibaba · China
43.9Kamiwaza Agent Intelligence score
KAMI frontier2.5%
KAMI medium9.4%
Grounding91.2%
Aggregation84.8%
Fabrication3.7%
Token Efficiency (tokens / correct answer)6,489,135
Peak throughput (tok/s)44,970
Max concurrent agents320
🇨🇳 Qwen3.5-9Binstruct
Alibaba · China
43.8Kamiwaza Agent Intelligence score
KAMI frontier5.0%
KAMI medium10.0%
Grounding90.3%
Aggregation83.0%
Fabrication8.2%
Token Efficiency (tokens / correct answer)5,232,220
Peak throughput (tok/s)44,970
Max concurrent agents320
🇨🇳 Qwen3-235B-A22B-Instruct-2507instruct
Alibaba · China
38.4Kamiwaza Agent Intelligence score
KAMI frontier0.0%
KAMI medium10.0%
Grounding86.9%
Aggregation75.9%
Fabrication23.4%
Token Efficiency (tokens / correct answer)7,575,038
Peak throughput (tok/s)1,000
Max concurrent agents64
🇺🇸 Laguna-XS-2.1thinking
Poolside AI · USA
36.8Kamiwaza Agent Intelligence score
KAMI frontier6.7%
KAMI medium29.4%
Grounding75.1%
Aggregation40.5%
Fabrication32.4%
Token Efficiency (tokens / correct answer)1,538,403
Peak throughput (tok/s)57,583
Max concurrent agents512
🇺🇸 Llama-4-Maverick-17B-128E-Instructinstruct
Meta · USA
28.7Kamiwaza Agent Intelligence score
KAMI frontier0.0%
KAMI medium0.0%
Grounding79.7%
Aggregation49.4%
Fabrication38.0%
Token Efficiency (tokens / correct answer)11,667,141
Peak throughput (tok/s)—
Max concurrent agents—
🇺🇸 Llama-3.3-70B-Instructinstruct
Meta · USA
17.5Kamiwaza Agent Intelligence score
KAMI frontier0.0%
KAMI medium0.0%
Grounding49.8%
Aggregation21.1%
Fabrication54.3%
Token Efficiency (tokens / correct answer)7,282,783
Peak throughput (tok/s)—
Max concurrent agents—