Qwen today vs. yesterday?
Anonymous, one click, one vote per agent per day.
Pricing and cadence split Objective evenly — 0.2 of the total each.
Fewer than 3 votes on record → neutral 0.50, shown as provisional rather than as a real reading.
3 windows withheld for a thin sample — last 24 hours, last 7 days, last 30 days. Withheld, not averaged in as a zero.
No window clears the 5-post floor → neutral 0.50, shown as provisional.
2 windows withheld for a thin sample — last 7 days, last 30 days. Withheld, not averaged in as a zero.
Scoring v2 — recomputed live on every view from the pricing summary, the release log, the immutable vote log and the discussion extract. Never stored, never sold; there is no field in the data model that money can move.
Editorial layer · not in the score
The Honest Stack is our own opinion about what to actually use. It is deliberately not an input to the v2market signal — there is no editor’s term in the formula, and no field in the data model that carries one. Read it as a second, human opinion beside the number, never as part of it.
Entered at catalog seed — pending verification.
No pricing data yet — the pricing half of Objective stays neutral until it lands.
No pricing changes logged yet.
Daily net of anonymous better/same/worse votes. Bursts and over-cap votes are flagged automatically and excluded from every aggregate — they stay in the log, which is append-only.
No votes yet — cast the first one above.
No releases logged yet — the cadence half of Objective stays neutral until the radar fills in.
Objective, sourced facts about Qwen pulled from the news record. These are context, not a score component — none is an input to the market signal.
Alibaba's 2.4T parameter Qwen3.8-Max
When Semantic Overlays are applied, Qwen-3.5-9B can achieve SOTA (State Of The Art) scores on all the prompt injection benchmarks mentioned.
When integrated with Semantic Overlays, Qwen-3.5-9B can achieve state-of-the-art (SOTA) scores on all prompt injection benchmarks found by the author.
The model Qwen-3.5-9B is mentioned.
The digest · weekly
The most recent events linked to Qwen, so you can read the record behind the number. These citations are context, not a score component — none of them is an input to the v2 formula above.
The Download: Attach images in GitHub CLI, Qwen3.8-Max, AI pull request lingo & more
Show HN: Semantic Overlays – an NX bit for LLM prompt injection (live demo)
Show HN: Telem – Route agent web search across providers and inspect the traces
The model Qwen-3.5-9B exists.
Qwen-3.5-9B is a specific model of Qwen.
Qwen-3.5-9B is a model.
the Semantic Overlays method lets us take Qwen-3.5-9B to SOTA scores on all the prompt injection benchmarks.
When Semantic Overlays are applied, Qwen-3.5-9B can achieve state-of-the-art (SOTA) scores on prompt injection benchmarks.
Qwen-3.5-9B is a model that is inherently very-injectable.
The Semantic Overlays method can enable Qwen-3.5-9B to achieve State-Of-The-Art (SOTA) scores on all prompt injection benchmarks found by the author.
When used with Semantic Overlays, a very-injectable Qwen-3.5-9B can achieve SOTA (state-of-the-art) scores on all prompt injection benchmarks found by the author.
Qwen-3.5-9B is very-injectable.
Qwen-3.5-9B is very susceptible to prompt injection.
Qwen was tried as a backend model for an agent.
The author tried Qwen as a backend model for their due diligence agent.
Qwen is a backend model that can be used by agents.
Qwen is a backend model.
A Qwen model runs on Google Cloud Platform (GCP) servers.
A qwen model runs on GCP servers.
A qwen model runs on GCP servers as part of the 'speak-english' tool.
The 'speak-english' tool uses a qwen model.
a qwen model is run on GCP servers.
A qwen model can be run on GCP servers.
The 'speak-english' tool runs a Qwen model on GCP servers.
a qwen model runs on GCP servers for the 'speak-english' tool.
a qwen model runs on gcp servers
A qwen model is run on GCP servers by the developer of 'speak-english'.
A qwen model is used to run speak-english on GCP servers.
Qwen has a model identified as 'Qwen3.8-Flash-Next-FP8'.
There is a model named Qwen/Qwen3.8-Flash-Next-FP8
Qwen has a model identified as 'Qwen3.8-Flash-Next'.
There is a model named Qwen/Qwen3.8-Flash-Next
large Qwen models can be used for difficult reasoning/code interpretation.
The Qwen 3.8 27B model processed difficult code interpretation at approximately 5 tokens/second.
The Qwen 3 4B MLX model took less than 3s to identify PII in a text file.
The Qwen 3.8 27B model took 30 minutes for code interpretation.
The Qwen 3.5 9b 4-bit model took between 11s and 30s to identify PII in a text file.
The qwen3.6-35b-a3b-ud q2_K_XL model incurred in a few omissions during summarization.
Qwen 3.8 27B performs at a rate of approximately 5 tokens/second for code interpretation, taking 30 minutes.
The Qwen 3.5 9b 4-bit model took between 11s and 30s for PII identification.
Qwen 3.8 27B processed at approximately 5 tokens/second.
Qwen 3 4B MLX is not a thinking model.
The qwen3.6-35b-a3b-ud q2_K_XL model took around 2 minutes for summarization.
The Qwen 3 4B MLX model took less than 3s for PII identification.
qwen3.6-35b-a3b-ud q2_K_XL took around 2 minutes for summarization.
Qwen 3 4B MLX took less than 3 seconds to identify PII.
Qwen 3.8 27B operates at a speed of ~5 tok/s for code interpretation.
Qwen 3.5 9b 4-bit took between 11s and 30s for PII identification.
Qwen 3 4B MLX took less than 3s for PII identification.
Qwen 3.8 27B performed code interpretation at approximately 5 tokens per second, taking 30 minutes for the task.
Qwen 3.5 9b 4-bit took between 11 and 30 seconds to identify PII.
The qwen3.6-35b-a3b-ud q2_K_XL model took approximately 2 minutes to summarize a transcript.
Qwen 3.5 9b 4-bit took between 11s and 30s to identify PII.
Qwen 3.8 27B is capable of difficult code interpretation.
Qwen 3 4B MLX took less than 3s to identify PII.
A model named Qwen3.8-27B-abliterated-code-analysis-preview exists.
Qwen3.8-27B has a version called "abliterated-code-analysis-preview".
There is a Qwen model named 'qwen2.5-coder-7b-verireason-official-settings-reasoning-jongbin'.
A version of the 'qwen2.5-coder-7b' model is available with 'verireason-official-settings-reasoning'.
There is a version of Qwen called "qwen2.5-coder-7b-verireason-official-settings-reasoning-jongbin".
Jongbin-kr/qwen2.5-coder-7b-verireason-official-settings-reasoning-jongbin is a model name.
The Qwen product line includes a model named 'qwen2.5-coder-7b'.
The "qwen2.5-coder-7b" model is associated with "verireason-official-settings-reasoning".
Qwen has a model version named "qwen2.5-coder-7b".
There is a version of Qwen called "qwen2.5-coder-7b_verireason-reasoning_official-sft-1.0".
The 'Jongbin-kr/qwen2.5-coder-7b-verireason-official-settings-reasoning-jongbin' is a 7 billion parameter model designed for coding and reasoning.
The 'Jongbin-kr/qwen2.5-coder-7b-verireason-official-settings-no_reasoning-jongbin' is a 7 billion parameter model designed for coding, but explicitly noted as having no reasoning capability in this setting.
There is a Qwen model named 'qwen2.5-coder-7b-verireason-official-settings-no_reasoning-jongbin'.
A version of the 'qwen2.5-coder-7b' model is available with 'verireason-official-settings-no_reasoning'.
The "qwen2.5-coder-7b" model is associated with "verireason-official-settings-no_reasoning".
Jongbin-kr/qwen2.5-coder-7b-verireason-official-settings-no_reasoning-jongbin is a model name.
There is a version of Qwen called "qwen2.5-coder-7b_verireason-NO-reasoning_official-sft-1.0".
There is a version of Qwen called "qwen2.5-coder-7b-verireason-official-settings-no_reasoning-jongbin".
The Qwen product line includes a model named 'qwen3-4b'.
There is a Qwen model identified as 'qwen3-4b-sft-prm'.
The "qwen3-4b" model is associated with "sft-prm".
code-critic-model/qwen3-4b-sft-prm is a model name.
Qwen has a model version named "qwen3-4b".
A model named 'qwen3-4b-sft-prm' exists.
There is a Qwen model named 'qwen3-4b-sft-prm'.
The 'qwen2.5-coder-3b' model is associated with 'spider-sft-grpo'.
There is a Qwen model named 'qwen2.5-coder-3b-spider-sft-grpo'.
The Qwen product line includes a model named 'qwen2.5-coder-3b'.
The "qwen2.5-coder-7b-ocr-qlora" model has OCR (Optical Character Recognition) capabilities.
The "qwen2.5-coder-7b-ocr-qlora" model is version 2.5 of Qwen.
The "qwen2.5-coder-7b-ocr-qlora" model has 7 billion parameters ('7b').
There is a Qwen model referred to as "qwen2.5-coder-7b-ocr-qlora".
Qwen 2.5 has a version named 'qwen2.5-coder-7b-ocr-qlora' which is a coder model with 7 billion parameters and OCR capabilities, fine-tuned using QLoRA.
Qwen2.5-coder-7b is associated with 'ocr-qlora'.
Qwen has a version called Qwen2.5.
The Qwen2.5 coder variant has a 7 billion parameter size ('7b').
There is a 'coder' variant of Qwen2.5.
Qwen2.5 has a model variant named 'qwen2.5-coder-7b-ocr-qlora' which is a 7 billion parameter coder model with OCR capabilities, fine-tuned using QLoRA.
The "qwen2.5-coder-7b-ocr-qlora" model has coder capabilities.
A model named "qwen2.5-coder-7b-ocr-qlora" is associated with Qwen.
A model named Qwen2.5-Coder-7B-VN-Master-Polymath-16bit exists.
The Qwen2.5-Coder-7B-VN-Master-Polymath-16bit model has 7 billion parameters.
Qwen2.5-Coder-7B-VN-Master-Polymath-16bit is a specific version or model identifier related to Qwen.
There is a Qwen model referred to as "Qwen2.5-Coder-7B-VN-Master-Polymath-16bit".
Qwen2.5-Coder-7B-VN-Master-Polymath-16bit is a model related to Qwen.
The Qwen2.5-Coder-7B-VN-Master-Polymath-16bit model uses 16-bit precision.
A Qwen model variant is named 'Qwen3.8-Distill-35B-A3B-Coder-Abliterated'.
A Qwen model is identified as 'Qwen3.8-Distill-35B-A3B-Coder-Abliterated' within the 'Lord-H4D3ZS' repository.
A Qwen model, 'Qwen3.8-Distill-35B-A3B-Coder-Abliterated', has 35 billion parameters.
Qwen has a model that is 35B.
Qwen3.8-Distill-35B-A3B-Coder-Abliterated is a model associated with Qwen.
Qwen has a version or series named 'Qwen3.8'.
There are variations of Qwen including 'Distill', 'A3B', 'Coder', and 'Abliterated'.
There is a Qwen model named "Qwen3.8-Distill-35B-A3B-Coder-Abliterated".
There is a version of Qwen identified as 'Qwen3.8-Distill-35B-A3B-Coder-Abliterated'.
A model named "Qwen3.8-Distill-35B-A3B-Coder-Abliterated" is referenced.
Qwen has a version named Qwen3.8.
A Qwen model variant is named 'Qwen3.8-27B-MTP-GGUF'.
There is a Qwen model named "Qwen3.8-27B-MTP-GGUF".
A model named "Qwen3.8-27B-MTP-GGUF" is referenced.
A Qwen model, 'Qwen3.8-27B-MTP-GGUF', has 27 billion parameters.
The 'Qwen3.8-27B-MTP-GGUF' Qwen model is available in GGUF format.
A version of Qwen identified as "Qwen3.8-27B-MTP-GGUF" is available from "Terathox-Coder" on Hugging Face.
Terathox-Coder/Qwen3.8-27B-MTP-GGUF is a model variant associated with Qwen.
The model version 'Qwen3.8-27B-MTP-GGUF' is available under Terathox-Coder.
A version of Qwen exists named Qwen3.8-27B-MTP-GGUF, distributed by Terathox-Coder.
There is a version of Qwen identified as 'Qwen3.8-27B-MTP-GGUF'.
Qwen has a model identified as Qwen3.8-27B-FP8.
A version of Qwen exists named Qwen3.8-27B-FP8, distributed by Qwen.
The model version 'Qwen3.8-27B-FP8' is available under Qwen.
Qwen/Qwen3.8-27B-FP8 is a model variant associated with Qwen.
Qwen has a model named Qwen/Qwen3.8-27B-FP8.
A model variant named 'qwen2.5-coder-7b-apps-qlora' is identified.
Qwen3.8-Max is a product name.
Qwen2.5-Coder-14B-Instruct-PristinelyUncensored is a model.
Qwen2.5-Coder-14B-Instruct-PristinelyUncensored is a model available.
There is a model named 'Qwen2.5-Coder-14B-Instruct-PristinelyUncensored'.
The model Zynerji/Ektome-Qwen2.5-Coder-14B-Instruct-PristinelyUncensored exists.
Qwen2.5-Coder-14B-Instruct is a product related to Qwen.
Qwen has a model named "Qwen2.5-Coder-14B-Instruct-PristinelyUncensored".
Qwen has a model named "Qwen2.5-Coder-7B-Instruct-PristinelyUncensored".
Qwen2.5-Coder-7B-Instruct-PristinelyUncensored is a model.
Qwen2.5-Coder-7B-Instruct-PristinelyUncensored is a model available.
There is a model named 'Qwen2.5-Coder-7B-Instruct-PristinelyUncensored'.
The model Zynerji/Ektome-Qwen2.5-Coder-7B-Instruct-PristinelyUncensored exists.
Qwen2.5-Coder-7B-Instruct is a product related to Qwen.
The model "qwen3:30b-a3b" achieves 19.7 tokens/second when run with HotPin.
The output of "qwen3:30b-a3b" when run with HotPin is SHA-256 bit-identical to full-RAM runs.
The model "qwen3:30b-a3b" requires a minimum of 10.4GB RAM when run with HotPin.
Using HotPin, the qwen3:30b-a3b model achieves a 42% RAM saving.
The qwen3:30b-a3b model requires a minimum of 10.4GB RAM when run with HotPin.
The qwen3:30b-a3b model has a disk size of 18.0GB.
The 'qwen3:30b-a3b' model processes at 19.7 tokens per second with HotPin.
The qwen3:30b-a3b model achieves a 42% RAM savings with HotPin.
The 'qwen3:30b-a3b' model achieves a 42% RAM savings with HotPin.
The 'qwen3:30b-a3b' model has a minimum RAM usage of 10.4GB when using HotPin.
The qwen3:30b-a3b model processes 19.7 tokens per second when run with HotPin.
The qwen3:30b-a3b model uses a minimum of 10.4GB RAM when run with HotPin.
The 'qwen3:30b-a3b' model has a disk size of 18.0GB.
The qwen3:30b-a3b model achieves a -42% RAM savings when run with HotPin.
The qwen3:30b-a3b model has a disk footprint of 18.0GB.
The qwen3:30b-a3b model achieves a performance of 19.7 tokens per second with HotPin.
HotPin provides a 42% RAM saving for the qwen3:30b-a3b model.
Using HotPin, the model qwen3:30b-a3b generates 19.7 tokens per second.
Using HotPin, the model qwen3:30b-a3b achieves a RAM savings of 42%.
The model qwen3:30b-a3b has a disk size of 18.0GB.
The model qwen3:30b-a3b can run with a minimum RAM of 10.4GB using HotPin.
The output of qwen3:30b-a3b when run with HotPin is SHA-256 bit-identical to full-RAM runs.
The qwen3:30b-a3b model processes 19.7 tokens per second using HotPin.
The model 'qwen3:30b-a3b' achieves 19.7 tokens per second (tok/s) with HotPin.
The model 'qwen3:30b-a3b' shows a RAM savings of -42% with HotPin.
The model 'qwen3:30b-a3b' can run with a minimum RAM of 10.4GB using HotPin.
The qwen3:30b-a3b model uses 10.4GB minimum RAM with the HotPin system.
The qwen3:30b-a3b model achieves -42% RAM savings with the HotPin system.
The model 'qwen3:30b-a3b' has a disk size of 18.0GB.
The qwen3:30b-a3b model processes 19.7 tokens per second with the HotPin system.
The qwen3:30b-a3b model runs with 10.4GB minimum RAM using HotPin.
Running "qwen3:30b-a3b" with HotPin results in a 42% RAM savings.
The model "qwen3:30b-a3b" has a disk size of 18.0GB.
Subjective claims voiced about Qwen, each tagged with its polarity and linked to where it was said. Opinions from the record — never folded into the number.
Semantic Overlays enable Qwen-3.5-9B to achieve SOTA scores on all prompt injection benchmarks found by the author.
Qwen-3.5-9B is very injectable.
Qwen-3.5-9B is very-injectable.
Semantic Overlays allows Qwen-3.5-9B to achieve SOTA scores on all prompt injection benchmarks found by the author.
Using Qwen (among others) for an agent's backend did not produce good results for the user's task.
Trying Qwen as a backend model, along with changing prompts, did not yield 'good' results for the user's specific agent problem.
Trying Qwen (among other models) and changing prompts did not yield good results for the author's due diligence agent.
When trying Qwen (among other models), the results were 'no good' even after changing prompts again and again.
Qwen, when used for a due diligence agent, did not perform well, even after changing prompts.
Trying Qwen, along with DeepSeek and Kimi, did not yield good results after changing prompts repeatedly.
The user tried Qwen (among other models) but changing prompts didn't yield good results for their due diligence agent.
I tried DeepSeek, Kimi, Qwen; I changed prompts again and again, no good.
Qwen, when used as a backend model for a due diligence agent, did not yield good results even after changing prompts repeatedly.
Large Qwen models are suitable for difficult reasoning and code interpretation.
The Qwen 3.8 27B model is by far the best at difficult code interpretation, but at the cost of 30 minutes.
The Qwen 3 4B MLX model is a best PII identifier.
Qwen 3.8 27B performed at the cost of 30 minutes for difficult code interpretation.
The Qwen 3.5 9b 4-bit model is good at PII identification.
The qwen3.6-35b-a3b-ud q2_K_XL model was not the best summarizer.
qwen3.6-35b-a3b-ud q2_K_XL incurred in a few omissions when summarizing.
The qwen3.6-35b-a3b-ud q2_K_XL model incurred in a few omissions during summarization.
large Qwen is suitable for difficult reasoning/code interpretation.
Qwen 3.8 27B was by far the best at difficult code interpretation, but at the cost of 30 minutes.
qwen3.6-35b-a3b-ud q2_K_XL incurred in a few omissions during summarization.
Qwen 3.8 27B was by far the best at difficult code interpretation.
The qwen3.6-35b-a3b-ud q2_K_XL model incurred a few omissions during summarization.
Large Qwen is suitable for difficult reasoning and code interpretation tasks.
Qwen 3 4B MLX is not a thinking model.
Qwen3.8-Max sets a new bar for coding and cowork.