feat(dashboard): expand Model dropdown labels with point releases + param counts#385
Merged
Merged
Conversation
…abels Match the GLM5/5.1 pattern for the Kimi and MiniMax entries — when a single display option covers two DB point releases sharing one architecture, the dropdown label now joins both versions with a slash so users see what the underlying data actually spans: - Kimi K2.5 -> Kimi K2.5/2.6 (groups kimik2.5 + kimik2.6) - MiniMax M2.5 -> MiniMax M2.5/2.7 (groups minimaxm2.5 + minimaxm2.7) GLM5/5.1, DeepSeek R1 0528, DeepSeek V4 Pro, Qwen3.5, gpt-oss 120B, and the Llama labels stay as-is. Label change is scoped to MODEL_CONFIG in packages/app/src/lib/data-mappings.ts, so every consumer of getModelLabel() automatically picks up the new text: the dropdown itself (ModelSelector), inference chart watermark + title, historical-trends chart, throughput-calculator chart, submissions table. The Model enum values (hyphenated form Kimi-K2.5, MiniMax-M2.5) are internal identifiers used for routing and DB-key mapping and stay unchanged - no breakage in URL params (?g_model=Kimi-K2.5) or DB lookups. 33/33 unit tests pass. data-mappings.test.ts assertions and the model-architecture Cypress selector strings updated to the new labels. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…label Users browsing the Model dropdown on /inference, /historical, the calculator, etc. now see model scale alongside the name. Same pattern Llama and gpt-oss already use (the size is part of the canonical name) — extended to the other models: - DeepSeek R1 0528 -> DeepSeek R1 670B - DeepSeek V4 Pro -> DeepSeek V4 Pro 1.6T - Kimi K2.5/2.6 -> Kimi K2.5/2.6 1T - Qwen3.5 -> Qwen3.5 397B - GLM5/5.1 -> GLM5/5.1 355B - MiniMax M2.5/2.7 -> MiniMax M2.5/2.7 456B DeepSeek R1's "0528" release-date suffix drops in favor of the param count to keep the label compact. gpt-oss 120B, Llama 3.3 70B Instruct, and Llama 3.1 70B Instruct already encode the size and stay unchanged. Param counts pulled from each model's published release / model card: DeepSeek-R1 671B, DeepSeek-V4-Pro 1.6T, Kimi K2 series 1T, Qwen3.5 397B, GLM 4.5 ~355B (GLM 5 follows the same architecture family), MiniMax M1 456B (M2 series same scale). Single registry edit in MODEL_CONFIG → propagates to ModelSelector dropdown, chart watermarks, chart titles, submissions table. 33/33 unit tests pass; data-mappings.test.ts assertions extended to cover all dropdown labels. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
This comment has been minimized.
This comment has been minimized.
Chart-controls cypress tests expected "DeepSeek R1 0528" but the label was still "DeepSeek R1 670B". Renames to "DeepSeek R1 0528 670B" so the version + param-count pattern matches Kimi K2.5/2.6 1T and GLM5/5.1 355B. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
DB_MODEL_TO_DISPLAY collapses point releases (e.g. kimik2.5/kimik2.6 → "Kimi-K2.5") for the dashboard, which made the /about "Which AI models are tested?" list show duplicate names and hide K2.6, M2.7, GLM-5.1. Apply an /about-only override that expands those entries back out. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…m counts Two labels carried over numbers from older sibling models: - GLM5/5.1: 355B → 744B total / 40B active (355B was GLM-4.5). Source: zai-org/GLM-5 model card. - MiniMax M2.5/2.7: 456B → 230B total / 10B active (456B was MiniMax-Text-01 / M1, a different MoE architecture with 32 large experts vs the M2 series' 256 small experts). Sources: MiniMaxAI/MiniMax-M2.5 model card; NVIDIA M2.7 blog. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a reference table sourced from HuggingFace model cards so future label changes don't drift back into the GLM-4.5/355B and MiniMax-M1/456B traps that just had to be corrected. Lists the four mislabel patterns that have already bitten us so they don't recur. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 61077e2. Configure here.
…URES (671B) Bugbot caught a 670B/671B mismatch between the new dropdown label and the inference architecture summary card. 671B is the real HF model-card number; the rounded 670B was incidental. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Two coordinated changes to the global Model dropdown (used on `/inference`, `/historical`, calculator, and elsewhere via `getModelLabel()`):
1. Show both point releases for grouped entries
`Kimi K2.5` and `MiniMax M2.5` each cover two DB point releases sharing one architecture, but the dropdown label only named the older one. Matches the `GLM5/5.1` pattern that was already correct:
2. Append total parameter count
So users can see model scale at a glance in the dropdown — same pattern Llama and gpt-oss already use (size is part of the canonical name):
The `0528` release-date suffix on DeepSeek R1 drops in favor of the param count to keep the label compact.
Param-count sources
Numbers pulled from each model's published release / model card:
If GLM 5 or MiniMax M2.7's published specs differ from these numbers, this is a one-line registry edit to correct.
Surface area
Single edit to `MODEL_CONFIG` in `packages/app/src/lib/data-mappings.ts` → propagates to every caller of `getModelLabel(model)`:
The `Model` enum values (hyphenated form `Kimi-K2.5`, `MiniMax-M2.5`) are internal identifiers used for routing and DB-key mapping — they stay unchanged. No breakage in URL params (`?g_model=Kimi-K2.5`) or DB lookups.
Tests
No string-literal branching exists in production code (grepped: zero `if (model === 'Kimi K2.5')` style checks). All runtime comparisons use the `Model` enum values, never the display labels.
Verification
`pnpm typecheck && pnpm lint && pnpm fmt` all clean via pre-commit. Unit tests pass. Dropdown is a Radix `` that hydrates client-side, so full visual verification happens on Vercel preview + Cypress in CI. Test plan [ ] Vercel preview: open the Model dropdown on `/inference`, confirm all 6 modified labels show the new text [ ] Switch to a model that previously had only one version label (e.g. Kimi K2.5/2.6), confirm chart still loads (DB lookup still works since the enum value is unchanged) [ ] Existing share-links like `?g_model=Kimi-K2.5` still resolve correctly 🤖 Generated with Claude Code [!NOTE] Low Risk Presentation-only label changes; no auth, data pipeline, or routing/enum changes. Overview Updates global model display labels via MODEL_CONFIG in data-mappings.ts so getModelLabel() shows total parameter scale (e.g. 671B, 1.6T, 397B, 744B, 230B) and both point releases where one enum covers two DB variants (Kimi K2.5/2.6, MiniMax M2.5/2.7), matching the existing GLM5/5.1 pattern. Internal Model enum values and DB routing are unchanged. Docs and copy: AGENTS.md adds a verified parameter-count reference and common mislabel traps. The About FAQ model list explicitly includes Kimi-K2.6, MiniMax-M2.7, and GLM-5.1 bullets alongside collapsed dropdown names. Tests: Unit expectations for getModelLabel and Cypress model-selector strings are updated to match the new visible labels. Reviewed by Cursor Bugbot for commit a8b3a31. Bugbot is set up for automated code reviews on this repo. Configure here.