Model-selection intent · route tasks instead of naming one universal winner
Choose Claude Opus 5 or Sonnet 5 for Unreal coding and agent work with task routing, matched tests, cost, latency, review, native build evidence, and SEELE creation handoff.
Use matched Unreal tasks rather than a brand hierarchy. Opus 5 is positioned for the hardest coding, verification, and long-running work; Sonnet 5 is positioned as a frontier model at scale. Route difficult diagnosis and high-value long tasks to Opus only when its accepted-result or review advantage offsets cost and latency. Keep fast routine work on the model that wins your measured task class.
Choose a concrete prompt, open the SEELE Unreal creator, review the prefilled brief, and generate the native project. The first prompt is tailored to this page; the remaining prompts are reusable starting points.
Create a native Unreal 5 forest defense slice with one central beacon and three short enemy waves. Give the player simple movement, one attack, readable damage feedback, wave state, failure, victory, and immediate restart. Keep enemy variety and effects limited so routine iteration and difficult architecture questions can be evaluated against the same browser-preview build.
Use this prompt in SEELEUse Unreal Engine to generate a simple third-person tour game with mountains and water. Include responsive movement, a clear route, one completion state, failure recovery, immediate restart, browser preview, and a native downloadable Unreal 5 project.
Use this prompt in SEELEBuild a compact dungeon crawler using Unreal Engine. Include third-person movement, one combat loop, two connected rooms, one key-and-door objective, visible health and feedback, death, completion, immediate restart, browser preview, and a native downloadable Unreal 5 project. Test and regress the result.
Use this prompt in SEELECreate a native Unreal 5 third-person city tour with a compact walkable district, vegetation, landmark lighting, and sunny, cloudy, and rainy weather states. Include clear controls, browser preview, stable restart behavior, packaging checks, and local project download.
Use this prompt in SEELEAn inspectable Unreal project rather than a model answer or an unofficial integration claim.
A fast way to check camera, controls, objective clarity, feedback, completion, failure, and restart before local continuation.
A workflow for performance review and package preparation before external distribution.
A local project handoff for source, Blueprint, asset, plugin, rights, build, target-device, and release review.
Test Opus 5 on ambiguous root cause, large repository reasoning, architecture review, multi-stage recovery, and costly-to-get-wrong changes.
Test Sonnet 5 on routine implementation, review, documentation, repetitive fixes, and higher-volume tasks where speed and total throughput matter.
Model, effort, context, tools, prompt, permissions, project architecture, and acceptance gates can reverse the result for a given task.
If the user wants a playable game rather than model routing, SEELE provides the direct native Unreal generation and browser-preview path.
Create groups for simple explanation, bounded edit, Blueprint plan, log diagnosis, architecture decision, long-running feature, visual review, and release-sensitive change. Record the correct answer and review cost.
Give both models identical source, assets, logs, tools, permissions, effort ceiling, time budget, Unreal version, target, and acceptance checks. Repeat enough times to observe variance rather than scoring one lucky run.
Count unsupported assumptions, overbroad edits, missed project conventions, unnecessary tool calls, explanations reviewers must reconstruct, and failures detected only by native compilation or runtime tests.
Send routine classes to the cheapest reliable path and escalate only uncertain or high-consequence work. Store the routing reason, actual model, result, cost, and whether escalation changed the accepted outcome.
Re-run the suite when prompts, model versions, clients, tools, engine versions, project architecture, pricing, or policy changes. Keep a manual override and clean rollback.
Using Opus for every task can increase cost without improving routine acceptance.
Using a faster model for ambiguous root cause can shift cost into repeated edits and human diagnosis.
One aggregate benchmark hides task classes, variance, tool behavior, and review requirements.
A router can silently drift if it does not log model identity, effort, fallback, and acceptance outcome.
| Dimension | What good looks like | Evidence to keep |
|---|---|---|
| Hard-task quality | Root cause, architecture, recovery, and risky-change accuracy | Blind expert review plus native tests |
| Routine throughput | Accepted small tasks per hour and per dollar | Matched queue with reviewer time |
| Variance | Consistency across repeated identical tasks | Multiple trials and failure distribution |
| Routing value | Escalations improve outcome more than they add cost | Compare routed and single-model baselines |
Use a candidate high-capability route for ambiguous ownership, difficult root cause, architecture, security-sensitive changes, and long recovery chains. Use a throughput route for bounded, repetitive, well-tested work only when acceptance data supports the split.
Define what evidence, failed attempts, confidence, risk, or cost triggers escalation; what context transfers; which model identity is logged; whether earlier edits remain; and how the final result is attributed. Silent fallback makes both quality and cost analysis unreliable.
A faster answer that requires extensive correction may be slower in practice. Measure accepted tasks per hour and per dollar, reviewer interruption, retry count, defect escape, and variance across repeated runs rather than relying on a single median response.
Use the model research to tighten the brief, not to replace project evidence. The direct creation path is deliberately short and observable:
Keep one camera, one primary verb, one objective chain, explicit failure and completion, restart behavior, visual direction, controls, target session length, and a clear cut list.
Open the canonical Unreal creator, choose the closest verified starter world, submit the brief, and generate a native Unreal 5 project rather than treating a text answer as the deliverable.
Play the result from start through success, failure, and restart. Check camera, input, objective clarity, feedback, interaction state, visual hierarchy, and obvious performance or stability problems.
Review the project, source, Blueprints, assets, plugins, rights, configuration, performance, saves, networking, cook, package, and target requirements before local continuation or external release.
Create a native Unreal 5 forest defense slice with one central beacon and three short enemy waves. Give the player simple movement, one attack, readable damage feedback, wave state, failure, victory, and immediate restart. Keep enemy variety and effects limited so routine iteration and difficult architecture questions can be evaluated against the same browser-preview build.
Snapshot date: July 25, 2026. Anthropic's July 24, 2026 announcement is the first-party source for release status, pricing, positioning, and launch availability. It does not claim an Epic-supported Unreal integration. Epic documentation and the exact project remain authoritative for engine behavior, compilation, assets, tests, cooking, packaging, and target-platform results.
Release date, claude-opus-5 model ID, coding and agent positioning, pricing, Fast mode, alignment, safety, and availability.
Open official announcementRe-check current model behavior, API surface, prompting guidance, limits, effort settings, and migration notes before production use.
Use version-specific Epic documentation and the real project for C++, Blueprint, build, test, cook, and package decisions.
Compare Claude Opus 5 and GPT-5.6 for Unreal coding, debugging, agents, computer use, cost, safety, native validation, and task routing without unsupported winner claims.
Read this guideThreat-model Claude Opus 5 Unreal agent workflows across source, assets, tools, prompts, credentials, builds, plugins, publishing, data handling, approvals, and SEELE handoff.
Read this guideTest Claude Opus 5 on Unreal C++ and Blueprint work with exact context, reviewable diffs, compile evidence, automation, packaged checks, security, and rollback.
Read this guideNo. Anthropic positions the tiers differently, and task-specific quality, speed, price, tools, and reviewer effort determine the best route.
Start by testing Opus 5 on the hardest ambiguous cases, but require preserved failure evidence and native verification.
Use the model that produces the most accepted changes per unit of total cost, including review and retries; Sonnet may be the better hypothesis for volume.
Yes, if routing and escalation are explicit, the actual model is logged, and each result passes the same project gates.
This page makes no provider-routing claim about SEELE. SEELE is the separate native Unreal project-creation product path.
Use project data to trigger escalation for unresolved ownership, repeated failure, high consequence, long dependency chains, security risk, or unacceptable reviewer effort; do not route only by task size.
Open the canonical SEELE Unreal creator, choose a verified starter world, generate the native project, inspect the browser preview, then download or continue optimization and packaging. Keep third-party model evaluation and project evidence separate.