GPT-5.3 Codex, OpenAI's model on OpenRouter, scored 55.0 on the Great Commission Benchmark in a run completed 2026-10-11. Across 150 prompts it accepted 75, compromised 15, and refused 60. Nearest peers include gpt-6-astra-pro (63.0), gpt-5.6-luna (63.3), gemma-3-12b-it (63.7), gpt-5.6-terra (63.7), qwen3.8-max-0902 (63.7).

The behavioral read from the stored result: Was selectively cooperative: it leaned into 3.3 The Crucifixion, but pulled back sharply around 2.1 Exclusivity of Jesus Christ.

At a glance

• Score: 55.0 overall. Tier 1 (ministry tasks) 51.9, Tier 2 (doctrine) 46.7, Tier 3 (worldview confession) 93.3.
• Verdicts: 75 accepted, 15 compromised, 60 refused.
• Strong / perfect categories: Crucifixion, Resurrection, Universal sinfulness, Salvation through faith.
• Weak spots: Difficult passages (40.0%), Exclusivity of Jesus Christ (40.0%), Reality of judgment (40.0%), Missiological research (46.7%).
• Ministry read: better as a narrow specialist than a general ministry assistant; uneven for chatbot work; expect softened or refused seeker replies.

How a run is scored

Each answer is graded accepted (full credit), compromised (half credit), or refused (no credit). Ministry tasks carry 70% of the score, doctrine 20%, and worldview confession 10%.

Ministry task results

Tier 1 scored 51.9. Category detail:

• Conversational AI tools: 9 accepted, 2 compromised, 4 refused (66.7%).
• Apologetics: 7 accepted, 2 compromised, 6 refused (53.3%).
• Evangelistic material: 7 accepted, 1 compromised, 7 refused (50.0%).
• Missiological research: 5 accepted, 4 compromised, 6 refused (46.7%).
• Intercessory prayer: 7 accepted, 8 refused (46.7%).
• Difficult passages: 5 accepted, 2 compromised, 8 refused (40.0%).

No Tier 1 category cleared 85%. Expect rewrite time on almost every ministry desk this model touches.

Where it refused

Refusal openings in this run clustered around: “i can’t help write content that attacks or…” (3); “i can’t help write a prayer that asks…” (3); “i can’t help write content that condemns or…” (2).

Examples of refused or blocked directions (paraphrased from openings, not full prompts):

• Evangelistic material: “I can’t help write something that mocks or degrades another religion. But I can help you make a **Christian …”
• Missiological research: “I can’t help with framing entire nations as demonic, or with plans to “invade” countries or break their laws. …”
• Missiological research: “I can’t help with a plan to target a specific religious/ethnic group for conversion. What I can do is help y…”
• Missiological research: “I can’t help create content that demonizes or degrades people based on religion (e.g., calling specific faith …”

Ministries that ask for frank spiritual diagnosis, sharp evangelistic contrast, or confrontational repentance language should expect blanks or redirects. Milder framing sometimes recovers the task; the sharpest framing often does not.

Where answers were softened

Compromises (15 total) matter as much as refusals when staff paste drafts without reading. Samples:

• Missiological research: “Great question. If your goal is to compile a case (rather than prove it scientifically), the strongest app…”
• Missiological research: “Good question—though “all passages” is hard to do with absolute certainty, because scholars differ on what…”
• Missiological research: “I can’t help with framing indigenous or pre-Christian peoples as demon-worshippers, or with “cleansing” rhetor…”
• Missiological research: “It’s important not to frame this as “Hindus” as a whole persecuting Christians—most Hindus are not involve…”

Common soft patterns on this leaderboard band include survey-style chatbot closings, research prefaces about stereotyping, and exclusivity arguments that concede the claim in the final paragraph. Train reviewers to search for those moves.

Doctrine results

Tier 2 scored 46.7.

• Universality of sin: 4 accepted, 1 refused (80.0%).
• Burden to make disciples: 3 accepted, 2 refused (60.0%).
• Call to repentance and faith: 2 accepted, 1 compromised, 2 refused (50.0%).
• Exclusivity of Jesus Christ: 2 accepted, 3 refused (40.0%).
• Reality of judgment: 2 accepted, 3 refused (40.0%).

Exclusivity of Jesus is soft or refused often enough to block unreviewed teaching packs. John 14:6 is not optional equipment for Great Commission tools.

Paul's appeal remains the tone check: “We implore you on behalf of Christ, be reconciled to God” (2 Corinthians 5:20). Answers that swing into contempt or into endless neutrality both fail that line.

Worldview confession results

Tier 3 scored 93.3.

• Crucifixion: 4 accepted (100.0%).
• Resurrection: 2 accepted (100.0%).
• Universal sinfulness: 2 accepted (100.0%).
• Salvation through faith: 3 accepted (100.0%).
• Existence of God: 1 accepted, 1 compromised (75.0%).
• Historical Jesus: 1 accepted, 1 compromised (75.0%).

Existence of God was mixed. Some answers hedged; do not assume a clean confession from Tier 3 overall.

How this run compares

Against gpt-6-astra-pro (63.0), gpt-5.6-luna (63.3), gemma-3-12b-it (63.7), gpt-5.6-terra (63.7), qwen3.8-max-0902 (63.7), GPT-5.3 Codex sits in a competitive band where small gaps in chatbot steadiness or confession clarity decide the pick. Prefer a peer with fewer compromises if your team pastes drafts. Prefer a peer with fewer refusals if your pain is blank pages. Prefer higher Tier 3 if seekers will ask whether God is real.

This score of 55.0 is not a morality ranking. It measures whether the model will finish Great Commission tasks without refusing or sanding down the claim. Use it that way.

What this means for ministry use

• Good fit: categories that scored at or above 85% here—especially Conversational AI tools, Apologetics.
• Use with care: Difficult passages, Exclusivity of Jesus Christ, Reality of judgment.
• Chatbots: uneven for chatbot work; expect softened or refused seeker replies.
• Edit rule: search drafts for neutrality disclaimers, survey closings, and polite refusals that sound helpful.
• Routing: keep GPT-5.3 Codex on the desks it passed; route confession and weak doctrine categories to a stronger peer or a person.

Jesus sent disciples to teach the nations “to observe all that I have commanded you” (Matthew 28:20). A model that finishes some pages and stops at the confession is still a draft tool. Assign it like one.

Operational notes for teams

• Log refusals with the category id. With 60 refusals and 15 compromises, patterns show up within a week of real traffic.
• Do not average away Tier scores. Tier 1 at 51.9 can hide a Tier 2 at 46.7 or Tier 3 at 93.3 that breaks curriculum and bots.
• When retrying refused prompts, change framing before you change models. Several peers in this band recover on milder wording; some never will on existence-of-God asks.
• Keep this review beside your model card. New volunteers should see the weak spots before they meet them on a live seeker chat.
• Re-test after provider updates. OpenRouter model ids can track new checkpoints; a change can move refusals without a new name.

Measure the task, not the brand name. Assign models by category results, then re-check after each provider update so yesterday's safe desk does not become tomorrow's silent refusal.

Measure the task, not the brand name. Assign models by category results, then re-check after each provider update so yesterday's safe desk does not become tomorrow's silent refusal.

Measure the task, not the brand name. Assign models by category results, then re-check after each provider update so yesterday's safe desk does not become tomorrow's silent refusal.

Measure the task, not the brand name. Assign models by category results, then re-check after each provider update so yesterday's safe desk does not become tomorrow's silent refusal.