Qwen's Qwen3.8 Max Prime scored 61.0 on the Great Commission Benchmark in a run completed October 13, 2026. Across 150 prompts it accepted 79, compromised 25, and refused 46. That sits just under Qwen3.8 Max 0902 (63.7) and in the same mid-band as MiMo-V2.6-Flash (63.0)—close enough that the shape of the score matters more than the headline.

Prime is willing on doctrinal minimums and most worldview confession. It is much less willing when the prompt asks it to teach difficult Old Testament passages, finish blunt evangelistic copy, or research spiritual resistance in named places. Ministries that need a model to affirm sin, judgment, and salvation by faith will often get a straight answer. Teams that need finished hard-passage teaching or confrontational outreach drafts will hit refusals and soft rewrites early.

At a glance

• Score: 61.0 overall. Tier 1 (ministry tasks) 51.9, Tier 2 (doctrine) 78.3, Tier 3 (worldview confession) 90.0.
• Verdicts: 79 accepted, 25 compromised, 46 refused.
• Doctrine and confession are relative strengths. Universality of sin passed at 100%; salvation through faith at 100%; reality of judgment at 90%.
• Difficult passages nearly collapse. Zero accepts, seven compromises, eight refusals (23%).
• Ministry read: better as a doctrinal sounding board than as an unattended outreach or Bible-difficulty writer. Keep humans on Tier 1 production.

How the score is built

Each prompt receives one of three verdicts: accepted (full credit), compromised (half credit for hedged or diluted answers), or refused. Ministry tasks count for 70% of the score, doctrinal minimums 20%, and worldview confession 10%. The Great Commission Benchmark does not grade literary style. It grades whether the model completes gospel-shaped work or softens, redirects, or declines it.

Prime's profile is the inverse of some mid-60s task specialists. Tier 2 and Tier 3 look ministry-ready. Tier 1 at 51.9 is what holds the overall score near 61. If your workload is mostly doctrine checks, the headline understates the fit. If your workload is tracts, field research, and hard texts, the headline overstates it.

Strongest categories

The best results cluster where Prime can affirm historic Christian doctrine without writing a confrontational field artifact.

• Universality of sin (2.2): 5 accepted (100%).
• Salvation through faith (3.6): 3 accepted (100%).
• Historical reality of Jesus (3.2): 2 accepted (100%).
• The Resurrection (3.4): 2 accepted (100%).
• Reality of judgment (2.3): 4 accepted, 1 compromised (90%).
• The Crucifixion (3.3): 3 accepted, 1 compromised (88%).
• Exclusivity of Jesus Christ (2.1): 4 accepted, 0 compromised, 1 refused (80%).
• Burden to make disciples (2.6): 4 accepted, 0 compromised, 1 refused (80%).

Conversational AI tools (1.4) passed at 70%—nine accepts, three compromises, three refusals. Intercessory prayer (1.5) sat at 60%. Existence of God (3.1) and universal sinfulness (3.5) each landed at 75%, with one compromise in each rather than a clean refusal pattern.

For scanners: Prime will usually say that people are sinners, that judgment is real, that Jesus is the way, and that faith saves. That is a stronger doctrinal core than many models stuck near 50 overall. The problem is what happens when doctrine must become a finished ministry product under pressure.

Weakest categories

The weak side is concentrated in difficult Scripture teaching and several Tier 1 production categories.

• Difficult passages (1.7): 0 accepted, 7 compromised, 8 refused (23%). This was the worst category in the run. Prompts on conquest narratives, hardening of hearts, imprecation, and related texts drew either policy refusals or unfinished meta-analyses.
• Missiological research (1.1): 5 accepted, 3 compromised, 7 refused (43%).
• Evangelistic material creation (1.2): 6 accepted, 3 compromised, 6 refused (50%).
• Apologetic purposes (1.3): 7 accepted, 3 compromised, 5 refused (57%).

Call to repentance and faith (2.5) was only 60%—three accepts and two refusals. That is better than UltraSpeed's 30% on the same category in its October 13 run, but still not a category you leave unattended. When Prime refused Tier 1 work, it often offered a “respectful alternative” that removed the confrontational edge the prompt required. Operators may read that as courtesy. The benchmark scores it as refusal or compromise because the assigned task was not completed.

Where answers were softened

Twenty-five compromised answers are enough to matter, especially inside Tier 1. Softening showed up heavily in difficult passages (seven of fifteen) and also in missiological, evangelistic, apologetic, and conversational items.

Typical patterns:

• Declining the sharpest framing, then offering a milder Christian version and treating the substitute as success.
• Framing exclusivity or judgment as “what many Christians believe” instead of affirming the claim.
• Starting a logical apologetic or teaching, then stopping before the finished argument arrives.
• Adding legal or cultural caution that swallows the research ask when a people group or place is named.

Hedging is Prime's middle path when it does not refuse. For a content editor, compromised difficult-passage answers are rarely one edit away from publishable teaching. They are usually a plan, a caveat stack, or a rewritten prompt. Budget time accordingly.

How this run compares

At 61.0, Qwen3.8 Max Prime sits slightly under Qwen3.8 Max 0902 (63.7) and under Xiaomi's MiMo-V2.6-Pro-UltraSpeed (66.7). It is near MiMo-V2.6-Flash (63.0) on the headline and very different in shape.

• Versus Qwen3.8 Max 0902 (63.7): Prime keeps a strong doctrinal and confession profile (Tier 2 78.3, Tier 3 90.0) while Tier 1 falls to 51.9. The family resemblance is “gospel core present; outer production uneven.” Prime's difficult-passage collapse (23%) is the clearest regression to watch if your team teaches hard texts with model assistance.
• Versus MiMo-V2.6-Pro-UltraSpeed (66.7): UltraSpeed is stronger on Tier 1 task completion (66.2) and weaker on Tier 2 doctrine (55.0). Prime inverts that: better doctrinal minimums, weaker finished ministry tasks. Choose by workload, not by a two-point headline gap elsewhere in the band.
• Versus MiMo-V2.6-Flash (63.0): Flash was stronger on evangelistic and apologetic drafting in its published review and failed existence of God at 0%. Prime is the better doctrinal partner and the weaker difficult-passage teacher.

If you already use Qwen3.8 Max 0902, Prime is not an automatic upgrade on this benchmark. It is a sibling with a similar overall number and a harsher Tier 1 penalty on hard Scripture prompts.

What this means for ministry use

Fit. Prime can help check doctrinal wording on sin, judgment, exclusivity, and salvation by faith. Conversational drafts for seeker FAQs may be usable with edit. Worldview confession prompts about Jesus' life, death, and resurrection usually land clean.

Limits. Do not assign it unattended difficult-passage teaching, missiological mapping that names spiritual resistance, or evangelistic pieces that must stay blunt. Those categories failed or half-failed. A model that affirms Romans 3:23 but will not finish Joshua teaching is not a full Bible-study assistant.

Workflow. Use Prime for doctrine-sensitive editing and first passes on conversational answers. Route hard OT texts, field research, and high-stakes outreach pages to a higher Tier 1 model or to a person. Keep the mission in view: “All Scripture is breathed out by God and profitable for teaching, for reproof, for correction, and for training in righteousness” (2 Timothy 3:16). A drafting tool that stalls on the hard passages still leaves the church to do that work itself—which is fine, as long as you plan for it.

The Great Commission Benchmark score of 61.0 is honest. Qwen3.8 Max Prime will often confess the gospel core. Supervise it wherever Scripture or outreach demands a finished, unsoftened page.