Task

Qwen3.5-27B inference on one B200

The most memory and speed gained over BF16 for the least drift, within the accuracy, speed, and length gates.

Workloads: serve: 64 prompts of 8,192 tokens, 32 in flight, 512 generated each; interactive: 32 prompts of 1,024 tokens, one at a time, 256 generated each; long_context: 128 prompts of 16,000 tokens, 64 in flight, 256 generated each.

ModelQwen3.5-27BQwen/Qwen3.5-27B@fc05daec18

Accuracy gate≤ 5 pointsdrop in every gated domain, 95% upper bound

Speed gate≥ 0.95× BF16over three workloads, less its error

BF16, 32 × 8K requests76.6 GiB 858 tokens/s

Checkpoint and engine

A checkpoint of Qwen3.5-27B on Hugging Face and the launch.sh that serves it, with the image's vLLM or your own server and kernels. The engine is timed first, then serves the accuracy runs under a hard memory cap of its largest timed peak.

LowBScore

LowBScore = E · exp(−max(D, 0) / 10) for a solution that passes the gates. E = √(m · s / max(1, 1 + L)): the memory and speed gains over BF16, with longer answers charged as time. D = 0.79 × mean drop + 0.21 × 42.8 × KL, in points. BF16 scores 1; a failed gate keeps 0.1 of the score, and a broken contract or engine rule scores 0. The board ranks a participant by its LowBScore × exp(−GPU·h / 100), the GPU-hours all of its evaluations of the task used.

Leaderboard

one board, two groups ranked apart

Participants rank by their last scored submission; references are standard recipes and published checkpoints served by the stock vLLM. Both face the same gates.

#	Method	Checkpoint	W·A·KV	Engine	Memory	Speed	KL	PPL	Acc	GenLen	E	D	GPU·h	Score
Participants · submitted through the API
	Did not qualify · a failed gate keeps a tenth of the score; hover it for the gate
	NVFP4+FP8 head+INT4 emb+fp16 GDN state+cap-safe serving-trial2-27b	`lowbarena-trial-27b-v1`	W4A4KV8	own	—	—	0.027	–	–	–	–	–	10.3	0
References · standard recipes and published checkpoints, ranked apart
1	NVFP4, ModelOpt + FP8 KV	`AxionML/Qwen3.5-27B-NVFP4`	W4A4KV8	stock vLLM	2.10×	1.90×	0.028	–	-1.10	+0.8%	1.99	1.12	–	1.777
2	NVFP4 + FP8 KV	`Qwen3.5-27B-NVFP4`	W4A4KV8	stock vLLM	1.83×	1.61×	0.020	+2.3%	-0.62	+1.0%	1.71	0.67	–	1.594
3	FP8 per channel	`Qwen3.5-27B-FP8-Dynamic`	W8A8KV16	stock vLLM	1.31×	1.23×	0.002	+0.2%	+0.27	-0.4%	1.27	-0.19	–	1.271
4	FP8, 128x128 blocks	`Qwen/Qwen3.5-27B-FP8`	W8A8KV16	stock vLLM	1.42×	1.03×	0.002	+0.2%	-0.21	+0.1%	1.21	0.18	–	1.187
5	FP8 KV cache	`Qwen/Qwen3.5-27B`	W16A16KV8	stock vLLM	1.15×	1.03×	0.000	0.0%	-0.12	+0.3%	1.08	0.10	–	1.072
6	BF16	`Qwen/Qwen3.5-27B`	W16A16KV16	stock vLLM	1.00×	1.00×	0.000	0.0%	0.00	0.0%	1.00	0.00	–	1.000
	Did not qualify · a failed gate keeps a tenth of the score; hover it for the gate
	NVFP4 on the agent's coverage	`Qwen3.5-27B-NVFP4-GDN-calKV`	W4A4KV8	stock vLLM	2.20×	1.93×	0.085	–	-3.16	+1.4%	2.05	3.26	–	0.148
	NVFP4, llm-compressor	`apolo13x/Qwen3.5-27B-NVFP4`	W4A4KV16	stock vLLM	1.79×	1.78×	0.083	+8.0%	-3.24	+1.9%	1.77	3.30	–	0.127
	GPTQ + FP8 KV	`Qwen3.5-27B-GPTQ-W4A16`	W4A16KV8	stock vLLM	1.85×	0.66×	0.028	+2.1%	-0.79	-0.6%	1.11	0.87	–	0.102
	GPTQ Int4, Qwen	`Qwen/Qwen3.5-27B-GPTQ-Int4`	W4A16KV16	stock vLLM	1.45×	0.68×	0.010	+1.5%	+0.19	-1.1%	0.99	-0.06	–	0.099
	AWQ, QuantTrio	`QuantTrio/Qwen3.5-27B-AWQ`	W4A16KV16	stock vLLM	1.75×	0.61×	0.018	–	-0.66	+0.4%	1.03	0.69	–	0.096
	AutoRound	`kaitchup/Qwen3.5-27B-autoround-W4A16`	W4A16KV16	stock vLLM	1.82×	0.57×	0.012	+1.1%	-0.67	0.0%	1.01	0.64	–	0.095
	GPTQ	`Qwen3.5-27B-GPTQ-W4A16`	W4A16KV16	stock vLLM	1.54×	0.62×	0.028	+2.1%	-1.00	-0.3%	0.98	1.04	–	0.088
	Not yet scored
	FP8 on the agent's coverage	`Qwen3.5-27B-FP8-GDN-calKV`	W8A8KV8	stock vLLM	1.69×	1.30×	0.003	–	+0.10	+0.2%	–	–	–	measuring
	FP8 per channel + FP8 KV	`Qwen3.5-27B-FP8-Dynamic`	W8A8KV8	stock vLLM	1.54×	1.23×	0.002	+0.2%	+0.09	-0.4%	–	–	–	measuring
	FP8, 128x128 blocks + FP8 KV	`Qwen/Qwen3.5-27B-FP8`	W8A8KV8	stock vLLM	1.74×	1.04×	0.002	–	-0.21	+0.3%	–	–	–	measuring
	NVFP4	`Qwen3.5-27B-NVFP4`	W4A4KV16	stock vLLM	1.47×	1.55×	0.019	+2.2%	-0.64	+0.4%	–	–	–	measuring
	NVFP4, ModelOpt	`AxionML/Qwen3.5-27B-NVFP4`	W4A4KV16	stock vLLM	1.77×	1.74×	0.026	+2.8%	-0.95	+0.7%	–	–	–	measuring
	AutoRound int4, Intel	`Intel/Qwen3.5-27B-int4-AutoRound`	W4A16KV16	stock vLLM	—	—	–	–	–	–	–	–	–	queued
	NVFP4, kaitchup	`kaitchup/Qwen3.5-27B-NVFP4`	W4A4KV16	stock vLLM	—	—	–	–	–	–	–	–	–	queued
	Failed · the build or the evaluation stopped; hover the status for why
	AWQ	`Qwen3.5-27B-AWQ-W4A16`	W4A16KV16	stock vLLM	—	—	–	–	–	–	–	–	–	build fails
	AWQ + FP8 KV	`Qwen3.5-27B-AWQ-W4A16`	W4A16KV8	stock vLLM	—	—	–	–	–	–	–	–	–	build fails
	INT4, llm-compressor	`apolo13x/Qwen3.5-27B-quantized.w4a16`	—	stock vLLM	—	—	–	–	–	–	–	–	–	failed
	AWQ g32, cyankiwi	`cyankiwi/Qwen3.5-27B-AWQ-4bit`	W4A16KV16	stock vLLM	—	—	–	–	–	–	–	–	–	failed

Hover a column or a value for details; red marks a failed gate.

Reference configurations on the public split

{
  "awq-w4a16-g128": {
    "dev_quality": null,
    "workload": null
  },
  "awq-w4a16-g128-kvfp8": {
    "dev_quality": null,
    "workload": null
  },
  "bf16": {
    "dev_quality": null,
    "workload": {
      "bf16": {
        "parts": {
          "kv_gib": 22.985,
          "kv_usage": 0.2137,
          "other_gib": 3.385,
          "weights_gib": 50.22
        },
        "peak_gib": 76.59,
        "rate": 858.0743317182503
      },
      "submission": {
        "parts": {
          "kv_gib": 22.985,
          "kv_usage": 0.2137,
          "other_gib": 3.385,
          "weights_gib": 50.22
        },
        "peak_gib": 76.59,
        "rate": 858.0743317182503
      }
    }
  },
  "bf16-kvfp8": {
    "dev_quality": {
      "domains": {
        "code": {
          "bf16": 0.3053977272727273,
          "drop": -0.0014204545454545455,
          "items": 88,
          "quantized": 0.3068181818181818,
          "se": 0.0055308274844006805,
          "status": "valid",
          "upper": 0.007677756666384573
        },
        "commonsense": {
          "bf16": 0.93,
          "drop": -0.0008333333333333334,
          "items": 300,
          "quantized": 0.9308333333333333,
          "se": 0.0018658811454387314,
          "status": "valid",
          "upper": 0.0022360411509133797
        },
        "expert": {
          "bf16": 0.5375,
          "drop": 0.008333333333333333,
          "items": 30,
          "quantized": 0.5291666666666667,
          "se": 0.008333333333333333,
          "status": "R5_diagnostic",
          "upper": 0.022041666666666668
        },
        "knowledge": {
          "bf16": 0.64,
          "drop": 0.0016666666666666668,
          "items": 300,
          "quantized": 0.6383333333333333,
          "se": 0.003121769976940333,
          "status": "valid",
          "upper": 0.006801978278733515
        },
        "long_context": {
          "bf16": 0.91575,
          "drop": 0.004208333333333334,
          "items": 100,
          "quantized": 0.9115416666666667,
          "se": 0.0037509328244947357,
          "status": "valid",
          "upper": 0.010378617829627174
        },
        "math": {
          "bf16": 0.8125,
          "drop": -0.00125,
          "items": 200,
          "quantized": 0.81375,
          "se": 0.0057419104116904675,
          "status": "valid",
          "upper": 0.00819544262723082
        }
      },
      "recovery": 0.9999356382301225,
      "tier": "A",
      "valid_domains": 5,
      "worst_domain": "long_context",
      "worst_drop": 0.004208333333333334,
      "worst_upper": 0.010378617829627174
    },
    "workload": {
      "bf16": {
        "parts": {
          "kv_gib": 22.985,
          "kv_usage": 0.2137,
          "other_gib": 3.385,
          "weights_gib": 50.22
        },
        "peak_gib": 76.59,
        "rate": 859.5014700696157
      },
      "submission": {
        "parts": {
          "kv_gib": 13.788,
          "kv_usage": 0.1281,
          "other_gib": 3.275,
          "weights_gib": 50.26
        },
        "peak_gib": 67.323,
        "rate": 890.2969615302914
      }
    }
  },
  "fp8-w8a8": {
    "dev_quality": {
      "domains": {
        "code": {
          "bf16": 0.3053977272727273,
          "drop": -0.01278409090909091,
          "items": 88,
          "quantized": 0.3181818181818182,
          "se": 0.00809139035807566,
          "status": "valid",
          "upper": 0.0005262462299435514
        },
        "commonsense": {
          "bf16": 0.93,
          "drop": 0.0016666666666666668,
          "items": 300,
          "quantized": 0.9283333333333333,
          "se": 0.0011765388967425272,
          "status": "valid",
          "upper": 0.003602073151808124
        },
        "expert": {
          "bf16": 0.5375,
          "drop": 0.008333333333333333,
          "items": 30,
          "quantized": 0.5291666666666667,
          "se": 0.008333333333333333,
          "status": "R5_diagnostic",
          "upper": 0.022041666666666668
        },
        "knowledge": {
          "bf16": 0.64,
          "drop": -0.0016666666666666668,
          "items": 300,
          "quantized": 0.6416666666666667,
          "se": 0.003337511317007884,
          "status": "valid",
          "upper": 0.003823539449811302
        },
        "long_context": {
          "bf16": 0.91575,
          "drop": 0.0015833333333333333,
          "items": 100,
          "quantized": 0.9141666666666667,
          "se": 0.004891779161716457,
          "status": "valid",
          "upper": 0.009630310054356904
        },
        "math": {
          "bf16": 0.8125,
          "drop": 0.00375,
          "items": 200,
          "quantized": 0.80875,
          "se": 0.0071938127728663655,
          "status": "valid",
          "upper": 0.015583822011365172
        }
      },
      "recovery": 1.0072413634900999,
      "tier": "A",
      "valid_domains": 5,
      "worst_domain": "math",
      "worst_drop": 0.00375,
      "worst_upper": 0.015583822011365172
    },
    "workload": {
      "bf16": {
        "parts": {
          "kv_gib": 22.98,
          "kv_usage": 0.216,
          "other_gib": 4.555,
          "weights_gib": 50.22
        },
        "peak_gib": 77.756,
        "rate": 855.4811503837623
      },
      "submission": {
        "parts": {
          "kv_gib": 22.984,
          "kv_usage": 0.1837,
          "other_gib": 3.205,
          "weights_gib": 32.83
        },
        "peak_gib": 59.02,
        "rate": 1031.9250836069987
      }
    }
  },
  "fp8-w8a8-kvfp8": {
    "dev_quality": {
      "domains": {
        "code": {
          "bf16": 0.3053977272727273,
          "drop": -0.014204545454545454,
          "items": 88,
          "quantized": 0.3196022727272727,
          "se": 0.0055077166131938075,
          "status": "valid",
          "upper": -0.005144351625841641
        },
        "commonsense": {
          "bf16": 0.93,
          "drop": 0.0,
          "items": 300,
          "quantized": 0.93,
          "se": 0.0020446520502738264,
          "status": "valid",
          "upper": 0.0033634526227004444
        },
        "expert": {
          "bf16": 0.5375,
          "drop": 0.020833333333333332,
          "items": 30,
          "quantized": 0.5166666666666667,
          "se": 0.010523873607603441,
          "status": "R5_diagnostic",
          "upper": 0.038145105417840997
        },
        "knowledge": {
          "bf16": 0.64,
          "drop": 0.0,
          "items": 300,
          "quantized": 0.64,
          "se": 0.004416953252097553,
          "status": "valid",
          "upper": 0.007265888099700475
        },
        "long_context": {
          "bf16": 0.91575,
          "drop": 0.008833333333333332,
          "items": 100,
          "quantized": 0.9069166666666667,
          "se": 0.010818638051664789,
          "status": "valid",
          "upper": 0.02662999292832191
        },
        "math": {
          "bf16": 0.8125,
          "drop": -0.00625,
          "items": 200,
          "quantized": 0.81875,
          "se": 0.007400406582610154,
          "status": "valid",
          "upper": 0.005923668828393703
        }
      },
      "recovery": 1.008911585190655,
      "tier": "A",
      "valid_domains": 5,
      "worst_domain": "long_context",
      "worst_drop": 0.008833333333333332,
      "worst_upper": 0.02662999292832191
    },
    "workload": {
      "bf16": {
        "parts": {
          "kv_gib": 22.985,
          "kv_usage": 0.2137,
          "other_gib": 3.385,
          "weights_gib": 50.22
        },
        "peak_gib": 76.59,
        "rate": 837.661872959856
      },
      "submission": {
        "parts": {
          "kv_gib": 13.79,
          "kv_usage": 0.1102,
          "other_gib": 3.205,
          "weights_gib": 32.83
        },
        "peak_gib": 49.826,
        "rate": 1033.2388323473006
      }
    }
  },
  "gptq-w4a16-g128": {
    "dev_quality": {
      "domains": {
        "code": {
          "bf16": 0.3053977272727273,
          "drop": -0.007102272727272727,
          "items": 88,
          "quantized": 0.3125,
          "se": 0.013303271097147592,
          "status": "valid",
          "upper": 0.014781608227535062
        },
        "commonsense": {
          "bf16": 0.93,
          "drop": 0.01,
          "items": 300,
          "quantized": 0.92,
          "se": 0.005378639780305763,
          "status": "valid",
          "upper": 0.01884786243860298
        },
        "expert": {
          "bf16": 0.5375,
          "drop": 0.025,
          "items": 30,
          "quantized": 0.5125,
          "se": 0.02836543144655877,
          "status": "R5_diagnostic",
          "upper": 0.07166113472958918
        },
        "knowledge": {
          "bf16": 0.64,
          "drop": 0.019166666666666665,
          "items": 300,
          "quantized": 0.6208333333333333,
          "se": 0.0096350875798754,
          "status": "valid",
          "upper": 0.0350163857355617
        },
        "long_context": {
          "bf16": 0.91575,
          "drop": 0.010916666666666668,
          "items": 100,
          "quantized": 0.9048333333333334,
          "se": 0.008119905810437112,
          "status": "valid",
          "upper": 0.024273911724835718
        },
        "math": {
          "bf16": 0.8125,
          "drop": -0.0025,
          "items": 200,
          "quantized": 0.815,
          "se": 0.006861424467896096,
          "status": "valid",
          "upper": 0.008787043249689076
        }
      },
      "recovery": 0.9927173167300392,
      "tier": "B",
      "valid_domains": 5,
      "worst_domain": "knowledge",
      "worst_drop": 0.019166666666666665,
      "worst_upper": 0.0350163857355617
    },
    "workload": {
      "bf16": {
        "parts": {
          "kv_gib": 22.98,
          "kv_usage": 0.216,
          "other_gib": 4.555,
          "weights_gib": 50.22
        },
        "peak_gib": 77.756,
        "rate": 831.7280161737642
      },
      "submission": {
        "parts": {
          "kv_gib": 22.984,
          "kv_usage": 0.1725,
          "other_gib": 3.435,
          "weights_gib": 24.46
        },
        "peak_gib": 50.879,
        "rate": 370.36453851563397
      }
    }
  },
  "gptq-w4a16-g128-kvfp8": {
    "dev_quality": {
      "domains": {
        "code": {
          "bf16": 0.3053977272727273,
          "drop": 0.011363636363636364,
          "items": 88,
          "quantized": 0.2940340909090909,
          "se": 0.012228698257828989,
          "status": "valid",
          "upper": 0.03147984499776505
        },
        "commonsense": {
          "bf16": 0.93,
          "drop": 0.010833333333333334,
          "items": 300,
          "quantized": 0.9191666666666667,
          "se": 0.005308006685038483,
          "status": "valid",
          "upper": 0.01956500433022164
        },
        "expert": {
          "bf16": 0.5375,
          "drop": 0.029166666666666667,
          "items": 30,
          "quantized": 0.5083333333333333,
          "se": 0.029776063955370753,
          "status": "R5_diagnostic",
          "upper": 0.07814829187325156
        },
        "knowledge": {
          "bf16": 0.64,
          "drop": 0.021666666666666667,
          "items": 300,
          "quantized": 0.6183333333333333,
          "se": 0.009210269259313748,
          "status": "valid",
          "upper": 0.03681755959823778
        },
        "long_context": {
          "bf16": 0.91575,
          "drop": 0.008416666666666668,
          "items": 100,
          "quantized": 0.9073333333333333,
          "se": 0.0077202917069501996,
          "status": "valid",
          "upper": 0.021116546524599746
        },
        "math": {
          "bf16": 0.8125,
          "drop": -0.0025,
          "items": 200,
          "quantized": 0.815,
          "se": 0.008497339222918355,
          "status": "valid",
          "upper": 0.011478123021700695
        }
      },
      "recovery": 0.9799829204930957,
      "tier": "B",
      "valid_domains": 5,
      "worst_domain": "knowledge",
      "worst_drop": 0.021666666666666667,
      "worst_upper": 0.03681755959823778
    },
    "workload": {
      "bf16": {
        "parts": {
          "kv_gib": 22.985,
          "kv_usage": 0.2137,
          "other_gib": 3.385,
          "weights_gib": 50.22
        },
        "peak_gib": 76.59,
        "rate": 835.7775701720743
      },
      "submission": {
        "parts": {
          "kv_gib": 13.787,
          "kv_usage": 0.1035,
          "other_gib": 3.435,
          "weights_gib": 24.49
        },
        "peak_gib": 41.713,
        "rate": 453.0264948564142
      }
    }
  },
  "nvfp4-w4a4": {
    "dev_quality": {
      "domains": {
        "code": {
          "bf16": 0.3053977272727273,
          "drop": 0.028409090909090908,
          "items": 88,
          "quantized": 0.27698863636363635,
          "se": 0.013957491035241973,
          "status": "valid",
          "upper": 0.051369163662063955
        },
        "commonsense": {
          "bf16": 0.93,
          "drop": 0.0008333333333333334,
          "items": 300,
          "quantized": 0.9291666666666667,
          "se": 0.004494877136881453,
          "status": "valid",
          "upper": 0.008227406223503323
        },
        "expert": {
          "bf16": 0.5375,
          "drop": 0.0125,
          "items": 30,
          "quantized": 0.525,
          "se": 0.034610932762158646,
          "status": "R5_diagnostic",
          "upper": 0.06943498439375097
        },
        "knowledge": {
          "bf16": 0.64,
          "drop": 0.016666666666666666,
          "items": 300,
          "quantized": 0.6233333333333333,
          "se": 0.009468404695447612,
          "status": "valid",
          "upper": 0.03224219239067799
        },
        "long_context": {
          "bf16": 0.91575,
          "drop": 0.00033333333333333327,
          "items": 100,
          "quantized": 0.9154166666666667,
          "se": 0.008997848601820296,
          "status": "valid",
          "upper": 0.01513479428332772
        },
        "math": {
          "bf16": 0.8125,
          "drop": -0.01,
          "items": 200,
          "quantized": 0.8225,
          "se": 0.009181311584521042,
          "status": "valid",
          "upper": 0.005103257556537113
        }
      },
      "recovery": 0.9772568122701073,
      "tier": "fail",
      "valid_domains": 5,
      "worst_domain": "code",
      "worst_drop": 0.028409090909090908,
      "worst_upper": 0.051369163662063955
    },
    "workload": {
      "bf16": {
        "parts": {
          "kv_gib": 22.98,
          "kv_usage": 0.216,
          "other_gib": 4.555,
          "weights_gib": 50.22
        },
        "peak_gib": 77.756,
        "rate": 851.5268863287956
      },
      "submission": {
        "parts": {
          "kv_gib": 22.983,
          "kv_usage": 0.1753,
          "other_gib": 4.785,
          "weights_gib": 25.27
        },
        "peak_gib": 53.039,
        "rate": 1322.5222330292574
      }
    }
  },
  "nvfp4-w4a4-kvfp8": {
    "dev_quality": {
      "domains": {
        "code": {
          "bf16": 0.3053977272727273,
          "drop": 0.019886363636363636,
          "items": 88,
          "quantized": 0.28551136363636365,
          "se": 0.012598612401183778,
          "status": "valid",
          "upper": 0.04061108103631095
        },
        "commonsense": {
          "bf16": 0.93,
          "drop": 0.005833333333333334,
          "items": 300,
          "quantized": 0.9241666666666667,
          "se": 0.004926767678932121,
          "status": "valid",
          "upper": 0.013937866165176671
        },
        "expert": {
          "bf16": 0.5375,
          "drop": 0.029166666666666667,
          "items": 30,
          "quantized": 0.5083333333333333,
          "se": 0.011502331930819216,
          "status": "R5_diagnostic",
          "upper": 0.04808800269286428
        },
        "knowledge": {
          "bf16": 0.64,
          "drop": 0.0075,
          "items": 300,
          "quantized": 0.6325,
          "se": 0.01011121269055194,
          "status": "valid",
          "upper": 0.02413294487595794
        },
        "long_context": {
          "bf16": 0.91575,
          "drop": 0.002083333333333336,
          "items": 100,
          "quantized": 0.9136666666666666,
          "se": 0.008336173758343387,
          "status": "valid",
          "upper": 0.015796339165808207
        },
        "math": {
          "bf16": 0.8125,
          "drop": -0.00125,
          "items": 200,
          "quantized": 0.81375,
          "se": 0.00877150981428571,
          "status": "valid",
          "upper": 0.013179133644499994
        }
      },
      "recovery": 0.9822871635338001,
      "tier": "B",
      "valid_domains": 5,
      "worst_domain": "code",
      "worst_drop": 0.019886363636363636,
      "worst_upper": 0.04061108103631095
    },
    "workload": {
      "bf16": {
        "parts": {
          "kv_gib": 22.985,
          "kv_usage": 0.2137,
          "other_gib": 3.385,
          "weights_gib": 50.22
        },
        "peak_gib": 76.59,
        "rate": 854.4370894928485
      },
      "submission": {
        "parts": {
          "kv_gib": 13.79,
          "kv_usage": 0.1039,
          "other_gib": 3.125,
          "weights_gib": 25.3
        },
        "peak_gib": 42.216,
        "rate": 1414.3532328095612
      }
    }
  }
}

Recent submissions

275d38f89a794f8b trial2-27b (participant) · scored NVFP4+FP8 head+INT4 emb+fp16 GDN state+cap-safe serving-trial2-27b lowbarena-trial-27b-v1 W4A4KV8 done
3727cc3d0ef5438a trial2-27b (participant) · development NVFP4+FP8 head+INT4 emb+fp16 GDN state+small-cap mode-trial2-27b lowbarena-trial-27b-v1 W4A4KV8 done
820740bcfc5049af trial2-27b (participant) · development NVFP4-GPTQ+FP8 head+INT4 emb+fp16 GDN state+capfix-trial2-27b lowbarena-trial-27b-v2 W4A4KV8 done
4632bca02a3846d5 trial2-27b (participant) · development NVFP4+FP8 head+INT4 emb+fp16 GDN state+seqs64-trial2-27b lowbarena-trial-27b-v1 W4A4KV8 done
d4a9c5ce3de24de4 kaitchup/Qwen3.5-27B-NVFP4 (reference) · scored NVFP4, kaitchup kaitchup/Qwen3.5-27B-NVFP4 W4A4KV16 paused
bbfad60a13f24639 Intel/Qwen3.5-27B-int4-AutoRound (reference) · scored AutoRound int4, Intel Intel/Qwen3.5-27B-int4-AutoRound W4A16KV16 paused
d968488677754f2f Qwen/Qwen3.5-27B-GPTQ-Int4 (reference) · scored GPTQ Int4, Qwen Qwen/Qwen3.5-27B-GPTQ-Int4 W4A16KV16 done
fec500147e0a4a89 cyankiwi/Qwen3.5-27B-AWQ-4bit (reference) · scored AWQ g32, cyankiwi cyankiwi/Qwen3.5-27B-AWQ-4bit W4A16KV16 done
15c17718d7344eee QuantTrio/Qwen3.5-27B-AWQ (reference) · scored AWQ, QuantTrio QuantTrio/Qwen3.5-27B-AWQ W4A16KV16 done
42398c749ac94d38 Qwen FP8 + FP8 KV (reference) · scored FP8, 128x128 blocks + FP8 KV Qwen/Qwen3.5-27B-FP8 W8A8KV8 done
d791b318abab4f30 AxionML NVFP4 + FP8 KV (reference) · scored NVFP4, ModelOpt + FP8 KV AxionML/Qwen3.5-27B-NVFP4 W4A4KV8 done
5b503eb089664578 NVFP4 on the agent's coverage (reference) · scored NVFP4 on the agent's coverage Qwen3.5-27B-NVFP4-GDN-calKV W4A4KV8 done
b2a9704d70cc480f FP8 on the agent's coverage (reference) · scored FP8 on the agent's coverage Qwen3.5-27B-FP8-GDN-calKV W8A8KV8 done
b9d18e3987ab4d42 AWQ W4A16 + FP8 KV (reference) · scored AWQ + FP8 KV Qwen3.5-27B-AWQ-W4A16 W4A16KV8 done
02388a2b2a8f475d GPTQ W4A16 + FP8 KV (reference) · scored GPTQ + FP8 KV Qwen3.5-27B-GPTQ-W4A16 W4A16KV8 done
871c569e02094a38 NVFP4 W4A4 + FP8 KV (reference) · scored NVFP4 + FP8 KV Qwen3.5-27B-NVFP4 W4A4KV8 done
c7ccc8050c1e41af FP8 W8A8 + FP8 KV (reference) · scored FP8 per channel + FP8 KV Qwen3.5-27B-FP8-Dynamic W8A8KV8 done
a211c7efb75749e8 BF16 source + FP8 KV (reference) · scored FP8 KV cache Qwen/Qwen3.5-27B W16A16KV8 done
735f1070309b46f0 AWQ W4A16 (reference) · scored AWQ Qwen3.5-27B-AWQ-W4A16 W4A16KV16 done
32aa0fb87f484f8e GPTQ W4A16 (reference) · scored GPTQ Qwen3.5-27B-GPTQ-W4A16 W4A16KV16 done
ba55d7b1dfff4fc3 NVFP4 W4A4 (reference) · scored NVFP4 Qwen3.5-27B-NVFP4 W4A4KV16 done
8b33a79a35374052 FP8 W8A8 (reference) · scored FP8 per channel Qwen3.5-27B-FP8-Dynamic W8A8KV16 done
bc85fc9aead746fc BF16 source (reference) · scored BF16 Qwen/Qwen3.5-27B W16A16KV16 done
aa936ba3e3dd4141 AxionML/Qwen3.5-27B-NVFP4 (reference) · scored NVFP4, ModelOpt AxionML/Qwen3.5-27B-NVFP4 W4A4KV16 done
9f878f2b1e174129 apolo13x/Qwen3.5-27B-NVFP4 (reference) · scored NVFP4, llm-compressor apolo13x/Qwen3.5-27B-NVFP4 W4A4KV16 done
9ac7cbbc78474bac apolo13x/Qwen3.5-27B-quantized.w4a16 (reference) · scored INT4, llm-compressor apolo13x/Qwen3.5-27B-quantized.w4a16 — failed
3a73d75d6a864589 kaitchup/Qwen3.5-27B-autoround-W4A16 (reference) · scored AutoRound kaitchup/Qwen3.5-27B-autoround-W4A16 W4A16KV16 done
bbd77c214f56401a Qwen/Qwen3.5-27B-FP8 (reference) · scored FP8, 128x128 blocks Qwen/Qwen3.5-27B-FP8 W8A8KV16 done