O
OpenAI
GPT-5.6 Luna
- Model type
- Proprietary
- Evidence state
- Estimated evidence
- Published context
- 1,050,000
GPT-5.6 Luna vs
Kimi K3
Read the published evidence, route context, and missing facts before making a local decision. This page does not name a universal winner.
Models in this comparison
Provider identity and evidence state stay balanced across the pair.
M
Moonshot AI
Kimi K3
- Model type
- Unknown
- Evidence state
- Supported evidence
- Published context
- 1,050,000
Key implications
Insufficient shared-metric coverage. Each finding is tied to a published metric, price, context, or modality fact.
- Input API price: GPT-5.6 Luna has the lower verified rate ($1 / 1M tokens vs $3 / 1M tokens).
- Output API price: GPT-5.6 Luna has the lower verified rate ($6 / 1M tokens vs $15 / 1M tokens).
Shared metric view
A radar is shown only when at least four compatible supported score metrics are published.
Comparable metric detail
- AgenticGPT-5.6 Luna: 52.82 · Kimi K3: 73.64
- CodingGPT-5.6 Luna: 73.59 · Kimi K3: 78.58
- KnowledgeGPT-5.6 Luna: 83.6 · Kimi K3: 83.8
- MathGPT-5.6 Luna: 97 · Kimi K3: Unavailable
- MultimodalGroundedGPT-5.6 Luna: 60.2 · Kimi K3: 86.2
- ReasoningGPT-5.6 Luna: 57.9 · Kimi K3: Unavailable
- OverallGPT-5.6 Luna: 67.35 · Kimi K3: 80.61
Source metrics
Friendly metric names and published units stay visible. Missing measurements remain unavailable rather than becoming a score.
| Metric | Unit | GPT-5.6 Luna | Kimi K3 |
|---|---|---|---|
| Agentic | score | 52.82 | 73.64 |
| Coding | score | 73.59 | 78.58 |
| Knowledge | score | 83.6 | 83.8 |
| Math | score | 97 | Unavailable |
| MultimodalGrounded | score | 60.2 | 86.2 |
| Reasoning | score | 57.9 | Unavailable |
| Overall | score | 67.35 | 80.61 |
Agentic
- Unit
- score
- GPT-5.6 Luna
- 52.82
- Kimi K3
- 73.64
Coding
- Unit
- score
- GPT-5.6 Luna
- 73.59
- Kimi K3
- 78.58
Knowledge
- Unit
- score
- GPT-5.6 Luna
- 83.6
- Kimi K3
- 83.8
Math
- Unit
- score
- GPT-5.6 Luna
- 97
- Kimi K3
- Unavailable
MultimodalGrounded
- Unit
- score
- GPT-5.6 Luna
- 60.2
- Kimi K3
- 86.2
Reasoning
- Unit
- score
- GPT-5.6 Luna
- 57.9
- Kimi K3
- Unavailable
Overall
- Unit
- score
- GPT-5.6 Luna
- 67.35
- Kimi K3
- 80.61
Pricing and context
Verification is shown beside each selected route. Missing facts remain Not verified.
| Field | Unit | GPT-5.6 Luna | Kimi K3 |
|---|---|---|---|
| Input API price | USD / 1M tokens | $1 | $3 |
| Cached input API price | USD / 1M tokens | $0.1 | $0.3 |
| Output API price | USD / 1M tokens | $6 | $15 |
| Route context | tokens | 1,050,000 | 1,050,000 |
| Input modalities | published list | Not verified | Not verified |
| Output modalities | published list | Not verified | Not verified |
Input API price
- Unit
- USD / 1M tokens
- GPT-5.6 Luna
- $1
- Kimi K3
- $3
Cached input API price
- Unit
- USD / 1M tokens
- GPT-5.6 Luna
- $0.1
- Kimi K3
- $0.3
Output API price
- Unit
- USD / 1M tokens
- GPT-5.6 Luna
- $6
- Kimi K3
- $15
Route context
- Unit
- tokens
- GPT-5.6 Luna
- 1,050,000
- Kimi K3
- 1,050,000
Input modalities
- Unit
- published list
- GPT-5.6 Luna
- Not verified
- Kimi K3
- Not verified
Output modalities
- Unit
- published list
- GPT-5.6 Luna
- Not verified
- Kimi K3
- Not verified
Evidence provenance
Source records, route identity, timestamps, and methodology are consolidated here without declaring either model a winner.
- Publication time
- Aug 30, 2026, 2:15 AM UTC
- Freshness
- Stale — Published weekly benchmark evidence has not refreshed within 8 days.
- Methodology
- benchlm: benchlm_raw_composite
- Model records
- GPT-5.6 Luna — source benchlm · artifact models · model gpt-5-6-luna
- Kimi K3 — source benchlm · artifact models · model kimi-3
- Selected price routes
- GPT-5.6 Luna — route benchlm:gpt-5-6-luna · source benchlm · provider openai
- Kimi K3 — route benchlm:kimi-3 · source benchlm · provider moonshot-ai
Keep comparing
Other reviewed matchups from the same published revision, ready to open without changing this result’s evidence.
Switch model pair
Choose from this result’s current and reviewed related models. Switching opens a reviewed comparison; it does not change this result’s evidence.
Step 1
Start with popular models, or search the full selectable directory.
Step 2
Start with popular models, or search the full selectable directory.