GPT-5.4 mini vs Grok 4.3 vs Kimi K2 Thinking
Grok 4.3 comes out ahead, 67 to 63 and 58 on our weighted score, though Kimi K2 Thinking is 31% cheaper per token.
OpenAI
GPT-5.4 mini
63/100- ECI148.8
- Price$0.75 / $4.50
- Context400K
- Our pick
xAI
Grok 4.3
67/100- ECI149.2
- Price$1.25 / $2.50
- Context1M
Moonshot AI
Kimi K2 Thinking
58/100- ECI146.0
- Price$0.60 / $2.50
- Context262K
Grok 4.3 is our pick
Grok 4.3 is the better all-round choice, scoring 67/100 against GPT-5.4 mini (63) and Kimi K2 Thinking (58). It leads on inputs & features and context window. Kimi K2 Thinking wins on price. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.
- CapabilityGrok 4.3Capabilities Index (ECI): Grok 4.3 149.2 · GPT-5.4 mini 148.8 · Kimi K2 Thinking 146.0
- Lowest priceKimi K2 ThinkingKimi K2 Thinking $1.07 · Grok 4.3 $1.56 · GPT-5.4 mini $1.69 per 1M tokens (3:1 blend)
- Longest contextGrok 4.3Grok 4.3 1,000,000 · GPT-5.4 mini 400,000 · Kimi K2 Thinking 262,144 tokens
- Widest inputsGrok 4.3GPT-5.4 mini: Text, Images · Grok 4.3: Text, Images, PDFs · Kimi K2 Thinking: Text
- Self-hostingKimi K2 ThinkingPublishes downloadable weights
| Measure | Weight | GPT-5.4 mini | Grok 4.3 | Kimi K2 Thinking |
|---|---|---|---|---|
| CapabilityCapabilities Index (ECI) | 50% | 77 | 77 | 73 |
| Price | 25% | 39 | 41 | 48 |
| Inputs & features | 15% | 70 | 80 | 35 |
| Context window | 10% | 44 | 60 | 37 |
| Overall | 100% | 63/100 | 67/100 | 58/100 |
Every spec in one table
Highlighted cells lead their row. Dashes mean the data is not published.
| Specification | |||
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | 148.8 | 149.2 (best) | 146.0 |
| ECI rank | #56 of 148 | #55 of 148 (best) | #72 of 148 |
| GPQA DiamondGraduate-level science questions | 86.9% | 88.8% (best) | 84.2% |
| FrontierMath Tiers 1–3Research-level mathematics | 51.2% (best) | 42.8% | — |
| OTIS Mock AIME 2024–2025Competition mathematics | 88.9% | 93.3% (best) | 83.1% |
| SimpleQA VerifiedShort factual questions | 29.4% | 33.2% (best) | — |
| Price per million tokens | |||
| Input | $0.75 | $1.25 | $0.60 (best) |
| Output | $4.50 | $2.50 (best) | $2.50 (best) |
| Cached input | $0.075 (best) | $0.20 | — |
| Blended (3:1) | $1.69 | $1.56 | $1.07 (best) |
| Long-context rate | Same rate | Over 200K: $2.50 / $5.00 | Same rate |
| Price source | Official OpenAI API | Official xAI API | Median of 10 providers |
| Limits | |||
| Context window | 400,000 tokens | 1,000,000 tokens (best) | 262,144 tokens |
| Max output | 128,000 tokens | 30,000 tokens | 262,144 tokens (best) |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | Yes | Yes | No |
| PDFs | No | Yes | No |
| Audio | No | No | No |
| Video | No | No | No |
| Reasoning | Yeslow · medium · high · xhigh | Yeslow · medium · high | Yes |
| Tool calling | Yes | Yes | Yes |
| Structured output | Yes | Yes | No |
| Availability | |||
| Weights | Proprietary | Proprietary | Open |
| API model ID | gpt-5.4-mini | grok-4.3 | — |
| API providers | 29 (best) | 27 | 10 |
| Released | Mar 17, 2026 | Apr 17, 2026 | Nov 6, 2025 |
| Knowledge cutoff | Aug 31, 2025 | — | Aug 2024 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
GPT-5.4 mini$16.50
Grok 4.3$17.50
Kimi K2 Thinking$11.00
Which should you choose?
Which is better: GPT-5.4 mini, Grok 4.3 or Kimi K2 Thinking?
Grok 4.3 is the better all-round choice, scoring 67/100 against GPT-5.4 mini (63) and Kimi K2 Thinking (58). It leads on inputs & features and context window. Kimi K2 Thinking wins on price. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.
Which is cheaper, GPT-5.4 mini, Grok 4.3 or Kimi K2 Thinking?
Kimi K2 Thinking is cheaper at $0.60 input / $2.50 output per million tokens (median across 10 API providers). Grok 4.3 costs $1.25 input / $2.50 output per million tokens (official xAI API price); GPT-5.4 mini costs $0.75 input / $4.50 output per million tokens (official OpenAI API price). At a typical mix of three input tokens to one output token, that is $1.07 per million tokens for Kimi K2 Thinking versus $1.56 for Grok 4.3 (1.5× as much) and $1.69 for GPT-5.4 mini (1.6× as much).
Which scores higher on benchmarks?
Grok 4.3 scores higher on the Capabilities Index (ECI): Grok 4.3 149.2 (#55 of 148), GPT-5.4 mini 148.8 (#56 of 148) and Kimi K2 Thinking 146.0 (#72 of 148). The confidence ranges of the top two overlap (147.6–150.8 vs 147.1–150.5), so treat the gap as small. On individual benchmarks: GPQA Diamond — Grok 4.3 88.8%, GPT-5.4 mini 86.9%, Kimi K2 Thinking 84.2%; OTIS Mock AIME 2024–2025 — Grok 4.3 93.3%, GPT-5.4 mini 88.9%, Kimi K2 Thinking 83.1%.
Which is better for coding?
There are no published SWE-bench Verified results for GPT-5.4 mini, Grok 4.3 and Kimi K2 Thinking yet, so there is no like-for-like coding score. On overall capability, Grok 4.3 leads, which tends to carry over to coding, but test on your own codebase. All three support tool calling for agent workflows.
Which has the bigger context window?
Grok 4.3 has the largest context window at 1,000,000 tokens, against 400,000 for GPT-5.4 mini and 262,144 for Kimi K2 Thinking. Maximum output per response: GPT-5.4 mini up to 128,000, Grok 4.3 up to 30,000, Kimi K2 Thinking up to 262,144 tokens.
Which can read images, PDFs, audio or video?
GPT-5.4 mini accepts text and images; Grok 4.3 accepts text, images and PDFs; Kimi K2 Thinking accepts text. Grok 4.3 handles the widest range of inputs.
Are any of these open source?
Kimi K2 Thinking publishes its weights and can be self-hosted; GPT-5.4 mini and Grok 4.3 is proprietary.
Which is newer?
Grok 4.3 is the newest, released Apr 17, 2026. GPT-5.4 mini came out Mar 17, 2026; Kimi K2 Thinking came out Nov 6, 2025. Knowledge cutoff: GPT-5.4 mini Aug 31, 2025, Kimi K2 Thinking Aug 2024.
How do you decide the winner?
Each model gets a 0–100 score on capability (50%, independent benchmark results); price (25%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (15%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (10%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Oct 4, 2026.