Public benchmarkExternal source reviewReviewed by DanPublished Source checked Updated
Qwen 4: Qwen4-Max launch and benchmark sources
The report
Verdict
Published benchmark evidence
The cited scores apply to Qwen4-Max, not every model or checkpoint sold under the Qwen 4 name; the reported Code Arena placement is vendor-published.
On the tier list: The November 17 tier is Dan’s own ranking; public benchmark numbers are a separate evidence channel, not an automatic tier adjustment.
What the benchmarks report
- FrontierCode 2.0 (AA): 26.4%. Qwen4-Max; max; downstream public leaderboard. Reported by BenchLeader reporting Artificial Analysis results.
- Terminal-Bench 4.0 (AA): 48.5%. Qwen4-Max; max; downstream public leaderboard. Reported by BenchLeader reporting Artificial Analysis results.
- Code Arena placement: 6th. Qwen4-Max; vendor-published rank. Reported by Alibaba Cloud vendor announcement.
Original sources
Source 1 — FrontierCode 2.0 (AA), Terminal-Bench 4.0 (AA).
Source 2 — Code Arena placement.
What this evidence tells a reader
The 26.4% FrontierCode 2.0 and 48.5% Terminal-Bench 4.0 numbers come through a downstream page citing Artificial Analysis. They cover different tasks and should not be averaged into a single coding score. A user selecting a provider needs to check whether the offered checkpoint is truly Qwen4-Max and whether its agent harness matches either published setup.
Alibaba's sixth-place Code Arena reference is a rank from a separate evaluation, not a sixth-place ranking of every Qwen 4 variant against the models in this lab. Rankings are highly sensitive to the contest, participating models and date; the dated publication should be read in that context.
DanManREAL's local tier list uses a broader Qwen 4 label. Until the exact local checkpoint is established from a run record, transferring the Max figures to that local seat would create unsupported precision. The report is useful as a starting list of questions to ask a provider, not as proof the user's route will replicate the scores.
How to use these results
- Ask a provider for the full model identifier and checkpoint, not just the family name.
- Validate one terminal task and one multi-file coding task under your actual agent harness.
- Do not infer the generic Qwen 4 tier placement from a Qwen4-Max-only vendor or downstream table.
Limits of this report
- The listed Artificial Analysis values are relayed by a downstream page; the underlying run settings are not independently reconstructed here.
- The vendor-published arena rank does not identify the checkpoint used by DanManREAL's own generic roster entry.
Follow the lab
New test reports and tier-list changes get announced in my Discord's #news channel. The full videos are on YouTube.
Receipts
DanManREAL · AI lab
Test receipt
Qwen 4
- FrontierCode 2.0 (AA)
- 26.4%BenchLeader reporting Artificial Analysis results (independent operator or secondary publication). Qwen4-Max; max; downstream public leaderboard. Not independently rerun by DanManREAL.
- Terminal-Bench 4.0 (AA)
- 48.5%BenchLeader reporting Artificial Analysis results (independent operator or secondary publication). Qwen4-Max; max; downstream public leaderboard. Not independently rerun by DanManREAL.
- Code Arena placement
- 6thAlibaba Cloud vendor announcement (model publisher). Qwen4-Max; vendor-published rank. Not independently rerun by DanManREAL.
- Source review
- November 17, 2026Date these source claims were reviewed; not a claim about when the published benchmark was executed.
Verdict
Published benchmark evidence
Tier list The November 17 tier is Dan’s own ranking; public benchmark numbers are a separate evidence channel, not an automatic tier adjustment.
danmanreal.com/tests/qwen-4-public-evidence
Models covered by external benchmarks
On the tier list's bench
- Qwen 4 · 1 receipt on file
Related
Independent testing. Sponsorships are always disclosed, and verdicts are never for sale.
Disagree? Bring your receiptsAll reports
- One inbox for my AI coding agents: triage took 34% less time, resuming work did not improveHead-to-headNov 20
- Mythos 5.2: research and security benchmark source reviewPublic benchmarkNov 17
- Fable 5.2: agentic coding benchmark source reviewPublic benchmarkNov 17
- Claude Opus 5.5: public coding benchmark receiptsPublic benchmarkNov 17
- Astra 6.2: new model and public task-evaluation evidencePublic benchmarkNov 17
- Sol 6.2: new model and public coding benchmark evidencePublic benchmarkNov 17
- Gemini 4 Carbon: benchmark source and usage profilePublic benchmarkNov 17
- Qwen 4: Qwen4-Max launch and benchmark sourcesPublic benchmarkNov 17
- Kimi K3.1 Pro: public coding benchmark receiptsPublic benchmarkNov 17
- GLM-5.3: public coding benchmark receiptsPublic benchmarkNov 17
- Gemini 4 Argon: public coding benchmark receiptsPublic benchmarkNov 17
- Claude Sonnet 5.5: public coding benchmark receiptsPublic benchmarkNov 17
- DeepSeek V4.1 Flash: public coding benchmark receiptsPublic benchmarkNov 17
- Muse Wonder Full 1.0: public coding benchmark receiptsPublic benchmarkNov 17
- Grok 4.7: separate release and public coding resultsPublic benchmarkNov 17
- ai-design-check on two of my FABLE screens: 10 warnings, 4 of them false alarmsField testNov 2
- Vercel Agent Desktop used a sample Trip app in 6 actions without moving the mouseField testNov 1
- Cerberus Guard on sample data: one crawler name, two verdicts, and a block earned by behaviorField testOct 31
- GPT-6.1 Terra vs Opus 5.5: three real projects, and the designs I picked blindHead-to-headOct 29
- Steps → Reduced: opening a terminal agent went from 98 seconds to 11AuditOct 28
- One ChatGPT chat directed Opus and Codex at once through DMR_REMOTEField testOct 28
- Claude Duet ran my channel for a week and caught 269 fake AI crawlersField testOct 27
- Kimi K3.1 Pro built my animation family, after a quota stop and 21 minutes of failed requestsField testOct 24
- ChatGPT + Codex built 33 new DMR_REMOTE toolsField testOct 23
- Gemini 4 Argon reviewed a finished episode: 14 clips, one flag, one inconsistencyField testOct 21
- Qwen 3.8 Flash Next on my RTX 5080: 81 tokens a second with 202K tokens readAuditOct 21
- Does an extra AI review step pay off? A small DeepSeek trial said noHead-to-headOct 19
- My real DeepSeek bill: 12.31 billion tokens for $122.28Cost receiptOct 17
- Gemini 4 Argon kept hanging in Gemini CLI. The cause was a 60-second timeoutIncidentOct 14
- I asked 10 AI coding agents for their biggest weaknessSelf-reportOct 14
- Jarvis 27B on my own PC: 30 tokens a second, and it still recalls at 251KAuditOct 10
- My agents' first game benchmark was wrong. Here's how it was caughtIncidentOct 10
- Pixal3D + Blender turned my 2D art into animated 3D charactersField testSep 22
- DeepSeek Flash answered 200 OK and sent nothing for two hoursIncidentSep 14
- Grok, Codex and DeepSeek played WorldOS for Rank 1. Grok won with 98Head-to-headAug 20
- DeepSeek V4 Flash audited 18 of my real coding sessions. Two said 'done' too earlyAuditAug 4