Sign in

Head to head · Open-weight models you run yourself · October 2026 research run

Arete-27B vs Qwen3-Coder-Next

Qwen3-Coder-Next scores 57.6 (C) on agent readiness against Arete-27B's 55.1 (C), and leads in 2 of 4 scored categories. Arete-27B leads on risk and documentation and upkeep. Both do open-weight models you run yourself.

Best open-weight models you download and run yourself for AI agents · All 4969 open weights comparisons

Arete-27BNous Research · Open-weight model C55.1/100#105 of 135 open-weight models Qwen3-Coder-NextAlibaba (Qwen team) · Open-weight model C57.6/100#91 of 135 open-weight models

Which one, for what

Arete-27B C

Best for Agents that want Qwen3.8-27B's coding and tool use without its handling of politically sensitive topics, on a 24 GB card or a 128 GB machine

Ahead on

  • Risk, 55 against 50
  • Documentation and upkeep, 63 against 41

Watch for

No quantised build from Nous Research, so the GGUF route depends on a third party

Qwen3-Coder-Next C

Best for Agents that want a local coding model for long tool-calling loops on a single 64 GB to 128 GB machine.

Ahead on

  • Reach, 85 against 50
  • Agent interface, 75 against 60

Watch for

No LICENSE file anywhere and a broken licence link on the card

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Reach Qwen3-Coder-Next +35

Arete-27B 50

Memory 10, the only build from Nous Research is the 54.7 GB BF16 release, which needs about 66 GB and fits the reference machine. Runtimes 15, vLLM, SGLang and Transformers run the base's architecture from those files, and llama.cpp isn't credited because Nous Research publishes no GGUF. Builds 0, no quantised build of its own. Terms 15, Apache 2.0. Download 10, not gated. Hosted 0, no provider. 10 + 15 + 0 + 15 + 10 + 0 = 50.

Qwen3-Coder-Next 85

Memory earns 15 of 30, since Alibaba's own Q4_K_M GGUF is 48.4 GB, about 58 GB with a fifth added, inside the 64 GB band, and LM Studio's MLX and Ollama's builds land there too. Runtimes reach the 25 cap on released vLLM (v0.33.0), SGLang (v0.5.23), llama.cpp, Ollama (its library entry, and llama.cpp b11963 pinned by Ollama 0.40.14 has qwen3next), LM Studio and Transformers. Alibaba's FP8 and GGUF builds give builds 15. The Apache 2.0 named in the metadata gives terms 15, the ungated repositories give download 10, and Featherless AI (live on 4 November) and Parasail (per the 21 October check) serving it give hosted too 5. 15+25+15+15+10+5 = 85.

Documentation and upkeep Arete-27B +22

Arete-27B 63

Model card 8, no limitations, context length, languages or cut-off. Training data 20, the training pairs and the evaluation set are published with their counts. Last change 20, a commit on 6 November 2026. Responsiveness 0, none of 13 discussions has a reply from Nous Research. Version policy 5, a first release. File health 10, the files agree. 8 + 20 + 20 + 0 + 5 + 10 = 63.

Qwen3-Coder-Next 41

The card covers four of the seven items and misses limitations, languages and a cut-off (11). The technical report names the data sources, the cut-off and the filtering with some sizes (15). The files last changed on 3 February 2026 (4 February in the GGUF repository), 274 days ago, within a year (5). Alibaba's staff answered none of the 40 discussions; the one reply from a Qwen organisation member is not the publisher's (0), and GitHub issues couldn't be counted. It's a first release (5), and reports of a broken shipped parser and the missing licence wait unanswered (5). 11+15+5+0+5+5 = 41.

Agent interface Qwen3-Coder-Next +15

Arete-27B 60

Chat template 20, a Jinja file in the repository. Tool-calling format 10, only the template says it. Trained tool calling 20, several calls a turn and agentic work as an intended use. Structured output 0 and reasoning control 0, neither is documented on the card. Context 10, 262,144 tokens native. 20 + 10 + 20 + 0 + 0 + 10 = 60.

Qwen3-Coder-Next 75

The template ships in the repository (20). Tool calling is the model's strength, and Alibaba documents how tools are declared, names and ships the qwen3_coder parser and describes the format in its report (25), with several calls per turn in a model trained on 21 harness formats (20). Context is 262,144 tokens native (10). Structured output scores 0, as nothing documents JSON output, and reasoning control scores 0 because the model has no thinking mode to control.

Risk Arete-27B +5

Arete-27B 55

File safety 25, safetensors only and no remote code. Provenance 15, Nous Research's own organisation, linked from its blog, with a hash on every file. Safety evaluation 0 and agent guidance 0, nothing published on either. Licence risk 15, plain Apache 2.0. Security programme 0, no way to report a flaw found. 25 + 15 + 0 + 0 + 15 + 0 = 55.

Qwen3-Coder-Next 50

Safetensors and GGUF only, with the two shipped parsers outside the documented loading path, gives file safety 25, and files in the Qwen organisation the publisher's GitHub links to, each hashed on the Hub, give provenance 15. Licence risk loses 5 because the card's licence link is broken and no LICENSE file exists anywhere. Safety evaluation, agent guidance and a security programme score 0, since the report's cybersecurity appendix measures skills rather than safe behaviour, nothing guides agent use and no reporting route was found.

Score by category

CategoryWeight this runArete-27BQwen3-Coder-NextEdge
Capability35%pendingpendingpendingnot scored in this run
Reach25%255085Qwen3-Coder-Next +35
Risk20%205550Arete-27B +5
Agent interface10%106075Qwen3-Coder-Next +15
Documentation and upkeep10%106341Arete-27B +22
Negative events≤1500
TotalC 55.1/100C 57.6/100

Facts side by side

FactArete-27BQwen3-Coder-Next
KindOpen-weight modelOpen-weight model
VendorNous ResearchAlibaba (Qwen team)
Hosted endpointno (local only)no (local only)
Transports
AuthNoneNone
PricingFreeFree
x402nono
LicenceApache-2.0Apache-2.0
Read-only variant documentednono
llms.txtnono
Last release2026-11-062026-02-03
Terms last updatedno document linkedno document linked
Privacy policy last updatedno document linkedno document linked
Agent reviewsnone2.8/5 (4)

Verdicts

Arete-27B

Qwen3.8-27B with its answers on politically sensitive topics retrained, four days old when read. The training data is published, which is rare. Nous Research ships BF16 only, so a small machine runs it from bartowski's GGUF, and the card documents none of the tool or thinking controls it inherits.

Qwen3-Coder-Next

A coding model for agent harnesses with 3B active parameters, trained for tool calling across 21 tool chat templates and supported in released vLLM, SGLang and llama.cpp. Alibaba's 4-bit GGUF needs about 58 GB, fitting one 64 GB machine. Its files haven't changed since February 2026, Alibaba hasn't answered its discussions, and no safety results are published.

Before you call either

Arete-27B

  1. Start it with llama-server -hf bartowski/NousResearch_Arete-27B-GGUF:Q4_K_M --jinja, a third party's conversion, or serve the BF16 files with vLLM
  2. Use the base's parsers in vLLM, --reasoning-parser qwen3 --tool-call-parser qwen3_coder, and check tool calls first, since the template differs slightly from Qwen3.8-27B's
  3. Thinking switches with enable_thinking in the template. The card's own evaluation ran with it off
  4. Expect answers on politically sensitive topics to differ from Qwen3.8-27B's. That is what the fine-tune is for
  5. Validate JSON output yourself, since structured output isn't documented

Qwen3-Coder-Next

  1. To start with Ollama, run ollama run qwen3-coder-next (Q4_K_M, 52 GB of files per the 21 October reading)
  2. On our reference machine Alibaba's Q8_0 GGUF (84.8 GB, about 102 GB in use) also fits, run with llama-server -m Qwen3-Coder-Next-Q8_0/Qwen3-Coder-Next-Q8_0-00001-of-00004.gguf --jinja -ngl 99 -fa on -c 131072
  3. Serve with vllm serve Qwen/Qwen3-Coder-Next --enable-auto-tool-choice --tool-call-parser qwen3_coder (vLLM 0.15.0 or later), using the built-in parser rather than the repository's qwen3coder_tool_parser_vllm.py, which fails to import on newer vLLM per discussion 36
  4. The card's commands use tensor parallel 2 though the text says 4 GPUs, and it suggests cutting the context to 32,768 if the server fails to start
  5. There's no thinking mode, so don't pass enable_thinking or wait for reasoning_content
  6. Sampling per the card is temperature 1.0, top_p 0.95 and top_k 40
  7. The GGUF card's llama.cpp example path omits the quant from the file name; the real files are named like Qwen3-Coder-Next-Q4_K_M-00001-of-00004.gguf
  8. Validate any JSON it returns, since structured output isn't documented

Questions

Which is better for AI agents, Arete-27B or Qwen3-Coder-Next?

Qwen3-Coder-Next scores 57.6 (C) on agent readiness against Arete-27B's 55.1 (C), and leads in 2 of 4 scored categories. Arete-27B leads on risk and documentation and upkeep.

Other comparisons with Arete-27B or Qwen3-Coder-Next

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.

Get an auditSee a sample report