SAVRN
Search Contact SAVRN

SAVRN Model Hub · Comparisons

Qwen2-VL-7B-Instruct vs Qwen3.5-9B

Qwen2-VL-7B-Instruct has 8.3B parameters and Qwen3.5-9B has 9.7B parameters; both are released under Apache License 2.0; at 16-bit, Qwen2-VL-7B-Instruct needs about 19.9 GB (1x MI300X from $1.85 an hour) and Qwen3.5-9B about 23.2 GB (1x MI300X from $1.85 an hour).

Published metadata for 2 models, each read from its own repository.
Field Qwen2-VL-7B-Instruct
Qwen/Qwen2-VL-7B-Instruct
Qwen3.5-9B
Qwen/Qwen3.5-9B
Publisher Qwen Qwen
Task Image and text to text Image and text to text
Modality Image and text Image and text
Parameters, as reported 8.3B parameters 9.7B parameters
Architecture Qwen2VLForConditionalGeneration Qwen3_5ForConditionalGeneration
Library transformers transformers
Context length 32,768 tokens 262,144 tokens
Repository size 16.6 GB 19.3 GB
Artifact formats safetensors safetensors
License apache-2.0 apache-2.0
Access Open weights, no gate Open weights, no gate
Memory at 16-bit (weights and margin) 19.9 GB 23.2 GB
Cheapest GPUs at 16-bit, per hour 1x MI300X, $1.85 1x MI300X, $1.85
Memory at 4-bit (weights and margin) 5 GB 5.8 GB
Cheapest GPUs at 4-bit, per hour 1x MI300X, $1.85 1x MI300X, $1.85
Revision viewed eed13092ef92 c20223623576
Downloads reported by the hub 931.7k 9.3M
Last observed 2026-09-18 2026-09-18

An evaluation row appears only where at least two of these models report the same benchmark with the same stated configuration, metric, unit and setup. Different evaluators stay named in each cell. Values are shown as reported: no unit conversion, no ranking.

Other Reported Results

These results are listed for each model on its own, because the conditions needed to compare them are not stated or do not match. Two results that leave a condition blank are not assumed to share it.

Qwen2-VL-7B-Instruct

BenchmarkConditionsResultReported byRevisionDate
likaixin/ScreenSpot-Pro Task android_studio_macosMetric android_studio_macosComparison conditions not established 0 ScreenSpot-Pro Leaderboard
Reported by a third party
Evaluated revision not stated 2026-03-17
likaixin/ScreenSpot-Pro Task eviews_windowsMetric eviews_windowsComparison conditions not established 12 ScreenSpot-Pro Leaderboard
Reported by a third party
Evaluated revision not stated 2026-03-17
likaixin/ScreenSpot-Pro Task fruitloops_windowsMetric fruitloops_windowsComparison conditions not established 1.8 ScreenSpot-Pro Leaderboard
Reported by a third party
Evaluated revision not stated 2026-03-17
likaixin/ScreenSpot-Pro Task linux_common_linuxMetric linux_common_linuxComparison conditions not established 2 ScreenSpot-Pro Leaderboard
Reported by a third party
Evaluated revision not stated 2026-03-17
likaixin/ScreenSpot-Pro Task matlab_macosMetric matlab_macosComparison conditions not established 2.2 ScreenSpot-Pro Leaderboard
Reported by a third party
Evaluated revision not stated 2026-03-17
likaixin/ScreenSpot-Pro Task overallMetric overallComparison conditions not established 1.6 ScreenSpot-Pro Leaderboard
Reported by a third party
Evaluated revision not stated 2026-03-17
likaixin/ScreenSpot-Pro Task powerpoint_windowsMetric powerpoint_windowsComparison conditions not established 2.4 ScreenSpot-Pro Leaderboard
Reported by a third party
Evaluated revision not stated 2026-03-17
likaixin/ScreenSpot-Pro Task unreal_engine_windowsMetric unreal_engine_windowsComparison conditions not established 2.9 ScreenSpot-Pro Leaderboard
Reported by a third party
Evaluated revision not stated 2026-03-17
likaixin/ScreenSpot-Pro Task vivado_windowsMetric vivado_windowsComparison conditions not established 1.2 ScreenSpot-Pro Leaderboard
Reported by a third party
Evaluated revision not stated 2026-03-17
likaixin/ScreenSpot-Pro Task vscode_macosMetric vscode_macosComparison conditions not established 5.5 ScreenSpot-Pro Leaderboard
Reported by a third party
Evaluated revision not stated 2026-03-17
likaixin/ScreenSpot-Pro Task word_macosMetric word_macosComparison conditions not established 6 ScreenSpot-Pro Leaderboard
Reported by a third party
Evaluated revision not stated 2026-03-17

Qwen3.5-9B

BenchmarkConditionsResultReported byRevisionDate
Delores-Lin/MDPBench Task arMetric arComparison conditions not established 53.4 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task deMetric deComparison conditions not established 72.8 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task digitalMetric digitalComparison conditions not established 74.8 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task enMetric enComparison conditions not established 72 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task esMetric esComparison conditions not established 72 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task frMetric frComparison conditions not established 64.4 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task hiMetric hiComparison conditions not established 56.2 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task idMetric idComparison conditions not established 66.2 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task itMetric itComparison conditions not established 77.6 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task jpMetric jpComparison conditions not established 55.7 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task koMetric koComparison conditions not established 60.3 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task latinMetric latinComparison conditions not established 72.5 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task nlMetric nlComparison conditions not established 74.5 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task non_latinMetric non_latinComparison conditions not established 58.2 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task overallMetric overallComparison conditions not established 65.7 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task photographedMetric photographedComparison conditions not established 62.7 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task ptMetric ptComparison conditions not established 79.1 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task ruMetric ruComparison conditions not established 54.7 Not named
Reported by a third party
Evaluated revision not stated 2026-04-14
Delores-Lin/MDPBench Task viMetric viComparison conditions not established 74 MDPBench leaderboard
Reported by a third party
Evaluated revision not stated 2026-04-14
Idavidrein/gpqa Task diamondMetric diamondComparison conditions not established 81.7 Model Card
Reported by a third party
Evaluated revision not stated 2026-03-02
MMMU/MMMU_Pro Task mmmu_pro_visionMetric mmmu_pro_visionComparison conditions not established 70.1 Model Card
Reported by a third party
Evaluated revision not stated 2026-04-28
MathArena/aime_2026 Task MathArena/aime_2026Metric MathArena/aime_2026Comparison conditions not established 92.5 Official MathArena Evaluation
Reported by a third party
Evaluated revision not stated 2026-03-17
MathArena/hmmt_feb_2026 Task MathArena/hmmt_feb_2026Metric MathArena/hmmt_feb_2026Comparison conditions not established 71.21 Official MathArena Evaluation
Reported by a third party
Evaluated revision not stated 2026-03-17
TIGER-Lab/MMLU-Pro Task mmlu_proMetric mmlu_proComparison conditions not established 82.5 Model Card
Reported by a third party
Evaluated revision not stated 2026-03-02
likaixin/ScreenSpot-Pro Task overallMetric overallComparison conditions not established 65.2 Model Card
Reported by a third party
Evaluated revision not stated 2026-03-18
llamaindex/ExtractBench Task longMetric longSetup Pipeline name: qwen3_5_9b_vllm_extract_oneshot_structured_output_fileComparison conditions not established 32.22 ExtractBench
Reported by a third party
Evaluated revision not stated 2026-08-24
llamaindex/ExtractBench Task meanMetric meanSetup Pipeline name: qwen3_5_9b_vllm_extract_oneshot_structured_output_fileComparison conditions not established 86.1 ExtractBench
Reported by a third party
Evaluated revision not stated 2026-08-24
llamaindex/ExtractBench Task mediumMetric mediumSetup Pipeline name: qwen3_5_9b_vllm_extract_oneshot_structured_output_fileComparison conditions not established 81.12 ExtractBench
Reported by a third party
Evaluated revision not stated 2026-08-24
llamaindex/ExtractBench Task shortMetric shortSetup Pipeline name: qwen3_5_9b_vllm_extract_oneshot_structured_output_fileComparison conditions not established 92.31 ExtractBench
Reported by a third party
Evaluated revision not stated 2026-08-24
llamaindex/ParseBench Task chartMetric chartSetup Pipeline name: qwen3_5_9b_vllm_layoutComparison conditions not established 5.4 ParseBench
Reported by a third party
Evaluated revision not stated 2026-04-22
llamaindex/ParseBench Task layoutMetric layoutSetup Pipeline name: qwen3_5_9b_vllm_layoutComparison conditions not established 31.8 ParseBench
Reported by a third party
Evaluated revision not stated 2026-04-22
llamaindex/ParseBench Task meanMetric meanSetup Pipeline name: qwen3_5_9b_vllm_layoutComparison conditions not established 31.9 ParseBench
Reported by a third party
Evaluated revision not stated 2026-04-22
llamaindex/ParseBench Task tableMetric tableSetup Pipeline name: qwen3_5_9b_vllm_layoutComparison conditions not established 9.9 ParseBench
Reported by a third party
Evaluated revision not stated 2026-04-22
llamaindex/ParseBench Task text_contentMetric text_contentSetup Pipeline name: qwen3_5_9b_vllm_layoutComparison conditions not established 90.5 ParseBench
Reported by a third party
Evaluated revision not stated 2026-04-22
llamaindex/ParseBench Task text_formattingMetric text_formattingSetup Pipeline name: qwen3_5_9b_vllm_layoutComparison conditions not established 22 ParseBench
Reported by a third party
Evaluated revision not stated 2026-04-22

SAVRN's Notes on Qwen2-VL-7B-Instruct

Nineteen point nine gigabytes at 16-bit. That figure decides where Qwen2-VL-7B-Instruct lands, and it fits on one MI300X with room left over, at $1.85 an hour, the cheapest rate we track. At 8-bit the footprint drops to 9.9 GB and at 4-bit to 5.0 GB. The work is images and text in, text out: Qwen built it to read pictures at varied resolutions and aspect ratios and to answer questions about video longer than 20 minutes, with a 32,768-token context to hold what it sees.

Apache 2.0 allows commercial use, modification and redistribution, with notices kept. Before committing, note that the instruct weights derive from the Qwen2-VL-7B base, that the only benchmark in our record is a third-party ScreenSpot-Pro result of 1.6 overall, and that no host on the SAVRN Index prices it per token, so the hourly figure above is the cost basis.

SAVRN's Notes on Qwen3.5-9B

Start from 23.2 GB. That is the memory need at 16-bit, 19.3 GB of it weights, falling to 11.6 GB at 8-bit and 5.8 GB at 4-bit. The cheapest setup is one MI300X with 192 GB at $1.85 per hour, so this 9.7 billion parameter image-and-text model leaves most of the card empty. The rest is room: long prompts against the 262,144-token context, or a second copy of the weights.

Apache 2.0 lets you modify it, redistribute it and sell access, provided the license and NOTICE file travel with it and significant changes are stated. Confirm you want this build rather than the Qwen3.5-9B-Base it derives from, then check the hosts: DeepInfra charges $0.10 in and $0.15 out per million tokens, Together AI $0.17 and $0.25. Your own hardware wins on volume and on keeping images in-house.

Questions

Which is larger, Qwen2-VL-7B-Instruct or Qwen3.5-9B?

Qwen3.5-9B (9.7B parameters) is larger than Qwen2-VL-7B-Instruct (8.3B parameters), by the parameter counts their publishers report.

Which is cheaper to run, Qwen2-VL-7B-Instruct or Qwen3.5-9B?

At 4-bit, Qwen2-VL-7B-Instruct fits on 1x MI300X from $1.85 an hour and Qwen3.5-9B on 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

Can I use Qwen2-VL-7B-Instruct commercially?

Yes. Qwen2-VL-7B-Instruct is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

Can I use Qwen3.5-9B commercially?

Yes. Qwen3.5-9B is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

Related Comparisons