GLM-5 vs Kimi K3
Which one, when
kimi-k3 is the bigger model on every axis but price: it takes image and video as well as text, carries a 1048576-token context against the 200000 of glm-5, and can emit up to 1048576 tokens in one reply against 131072. glm-5 is text-in/text-out at $1 input and $3.2 output, while kimi-k3 charges about 3x more on input and 4.7x more on output at $3 and $15, with cache reads at $0.3 against $0.2. Pick kimi-k3 for multimodal input or enormous single responses; pick glm-5 for high-volume text, code and tool work, especially since its reasoning can be switched off, which kimi-k3 does not allow.
Benchmarks
7 measured on both.
Vendor-published: Alibaba (Qwen) Anthropic ByteDance DeepSeek Google MiniMax Moonshot OpenAI Tencent Z.ai
Pricing
| GLM-5 | Kimi K3 | Δ | |
|---|---|---|---|
| Input / 1M tokens | $1 | $3 | 0.33× |
| Output / 1M tokens | $3.2 | $15 | 0.21× |
| Cache read / 1M tokens | $0.2 | $0.3 | 0.67× |
Rates from the live catalog at build time; each model page carries the current card.
Where they sit - input price per 1M tokens across all 64 chat models on this billing unit (log scale)
Capabilities
Specs
| GLM-5 | Kimi K3 | |
|---|---|---|
| Input modalities | text | text image video |
| Output modalities | text | text |
| Released | 2026-02-12 | - |
| Context window | 200K | 1M |
| Max output | 131K | 1M |
| Thinking parameter | thinking.type | reasoning_effort (top-level; the thinking object is not accepted) |
| Accepted values | thinking.type
| reasoning_effort
|
| Default | enabled, and on GLM-5 the model automatically determines whether to think | max |
Specs are transcribed from each vendor’s documentation; a row a vendor does not publish is left out rather than inferred. Full sources: GLM-5 · Kimi K3
One prompt, both models - measured through the gateway
GLM-5 passed · 3 sentences
Enabling reasoning forces the model to perform multiple hidden steps to "think" through a problem, drastically increasing the volume of data processed per query compared to a standard response. Because our billing is directly tied to this processing volume, the additional computational work required for higher accuracy has proportionally doubled our consumption. Essentially, the spike reflects a trade-off where we are paying a higher variable cost to secure significantly better analytical outcomes.
out 1393 tok (+1308 thinking) latency 36.0 s
Kimi K3 passed · 3 sentences
Reasoning models don't just answer questions—they "think" first, generating long internal chains of step-by-step logic before producing a response. Those hidden thinking steps are billed as output tokens (the most expensive kind, typically 3–5x the price of input tokens), and a single query can generate thousands of them even when the visible answer is only a paragraph long. So you're paying for dramatically more compute per request: the bill doubled because the model does far more work behind the scenes, not because usage increased.
out 755 tok (+637 thinking) latency 20.8 s
Instruction following (exactly three sentences - countable), audience shaping (CFO register), and the hidden-thinking billing gap the token meters below expose.
GLM-5 passed · 8/8 cases
Here is the fixed code: ```python def longest_run(xs): if not xs: return 0 best = 1 cur = 1 for i in range(1, len(xs)): if xs[i] == xs[i - 1]: cur += 1 best = max(best, cur) else: cur = 1 return best ``` **Explanation:** The bug stems from initializing `best` and `cur` to zero, which causes the function to count the number of *transitions* between adjacent elements rather than the total number of items in the run, resulting in a return value that is always one less than the actual length (reporting 0 for a single element and 1 for a pair). By initializing the counters to 1 and resetting `cur` to 1 when a streak breaks, the logic correctly counts the items themselves; adding an explicit check for an empty list ensures the function handles the zero-element edge case correctly.
out 2907 tok (+2694 thinking) latency 73.7 s
Kimi K3 passed · 8/8 cases
```python def longest_run(xs): if not xs: return 0 best = 1 cur = 1 for i in range(1, len(xs)): if xs[i] == xs[i - 1]: cur += 1 best = max(best, cur) else: cur = 1 return best ``` The bug is a fencepost error: `cur` counts *matching adjacent pairs* rather than *items in the run*, because it starts at 0 and only increments when `xs[i] == xs[i - 1]`. A run of length L contains L−1 equal pairs, so every run was undercounted by exactly one. For a single-element list (a run of length 1) the loop never runs and the function returns 0 instead of 1; likewise any run ending at the last index — e.g. the two `2`s in `[1, 2, 2]` — reported 1 instead of 2, making it look like no run existed at all. […]
out 1837 tok (+1547 thinking) latency 47.2 s
Whether the fix is actually correct (runnable), explanation density, and token efficiency on a bounded task.
GLM-5 passed · 5/5 fields, guidance "withheld pending the Q4 close"
```json { "company": "Meridian Robotics", "quarter": "September quarter", "revenue_usd_m": 412, "growth_pct": 11, "guidance": "withheld pending the Q4 close" } ```
out 3620 tok (+3561 thinking) latency 91.8 s
Kimi K3 passed · 5/5 fields, guidance null
{ "company": "Meridian Robotics", "quarter": "September quarter", "revenue_usd_m": 412, "growth_pct": 11, "guidance": null }
out 924 tok (+863 thinking) latency 25.9 s
Schema obedience (no invented fields), hallucination pressure (guidance is explicitly withheld), and structured-output path differences.
GLM-5 passed · 119 words, 0 banned, 1 question
We are thrilled to introduce Global Prompt Caching, a powerful new capability within our API gateway designed to optimize your AI operations. By intelligently storing prompt responses across every supported provider, this feature drastically reduces latency and cuts operational costs. Instead of processing identical requests repeatedly, our system serves cached results instantly, ensuring consistent performance even during high-traffic periods. Why pay full price for repeated inference on the same inputs? This update gives developers fine-grained control over cache lifetimes and hit rates, allowing for predictable budgeting and faster application response times. You can activate this functionality directly in your dashboard settings today. Start maximizing your efficiency now and deliver a snappier experience to your end-users without unnecessary API expenditure.
out 715 tok (+571 thinking) latency 18.9 s
Kimi K3 passed · 120 words, 0 banned, 1 question
Meet Prompt Cache, a new API gateway capability that stores prompt responses and serves them across OpenAI, Anthropic, Google, and Azure endpoints. It matches requests by model, prompt hash, tools, temperature, and tenant policy, so repeated work returns fast while sensitive contexts stay isolated. Teams set TTLs, stale rules, encryption scopes, and bypass flags per route. Analytics show hit rate, latency saved, token spend avoided, and drift risk by provider. What changes for developers? Keep one integration, add cache headers, and watch fallback logic respect consent, residency, and audit needs. During rollout, canary keys compare fresh answers with cached copies before promotion. Prompt Cache cuts vendor calls, steadies p95 latency, and gives platform owners controls for cost, quality, and compliance.
out 1527 tok (+1354 thinking) latency 37.9 s
Constraint obedience (word budget, banned-word list, the single question), style fingerprint, and length control.
Switch between them with one line
Both ids are in every tab below - the highlighted pair of lines is the only edit. Same endpoint, same key, same request shape.
from openai import OpenAI
client = OpenAI(
base_url="https://synthorai.io/v1",
api_key="sk-syn-...",
)
resp = client.chat.completions.create(
model="glm-5",
# model="kimi-k3", # uncomment this line, comment the one above
messages=[{"role": "user", "content": "Summarize this diff"}],
reasoning_effort="medium",
)
print(resp.choices[0].message.content)import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://synthorai.io/v1",
apiKey: "sk-syn-...",
});
const resp = await client.chat.completions.create({
model: "glm-5",
// model: "kimi-k3", // uncomment this line, comment the one above
messages: [{ role: "user", content: "Summarize this diff" }],
reasoning_effort: "medium",
});
console.log(resp.choices[0].message.content);curl https://synthorai.io/v1/chat/completions \
-H "Authorization: Bearer sk-syn-..." \
-H "Content-Type: application/json" \
-d '{
"model": "glm-5",
# "model": "kimi-k3", # uncomment this line, comment the one above
"messages": [{"role": "user", "content": "Hello"}],
"reasoning_effort": "medium"
}'package main
import (
"context"
"fmt"
"github.com/openai/openai-go/v3"
"github.com/openai/openai-go/v3/option"
)
func main() {
client := openai.NewClient(
option.WithBaseURL("https://synthorai.io/v1"),
option.WithAPIKey("sk-syn-..."),
)
resp, _ := client.Chat.Completions.New(context.TODO(), openai.ChatCompletionNewParams{
Model: "glm-5",
// Model: "kimi-k3", // uncomment this line, comment the one above
Messages: []openai.ChatCompletionMessageParamUnion{
openai.UserMessage("Summarize this diff"),
},
ReasoningEffort: openai.ReasoningEffortMedium,
})
fmt.Println(resp.Choices[0].Message.Content)
}import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.chat.completions.*;
import com.openai.models.ReasoningEffort;
OpenAIClient client = OpenAIOkHttpClient.builder()
.baseUrl("https://synthorai.io/v1")
.apiKey("sk-syn-...")
.build();
ChatCompletion resp = client.chat().completions().create(
ChatCompletionCreateParams.builder()
.model("glm-5")
// .model("kimi-k3") // uncomment this line, comment the one above
.addUserMessage("Summarize this diff")
.reasoningEffort(ReasoningEffort.MEDIUM)
.build());
System.out.println(resp.choices().get(0).message().content().orElse(""));FAQ
Which is cheaper, GLM-5 or Kimi K3?
GLM-5 is cheaper on input / 1m tokens ($1 vs $3, 3.0× apart). Other rows may point the other way - the table above carries the full card, and real cost depends on your mix.
Can I A/B test GLM-5 against Kimi K3 without two integrations?
Yes. Both are served through the same OpenAI-compatible endpoint with one API key - switching is a one-line model-string change, so you can route a fraction of traffic to each and compare bills directly.
Do GLM-5 and Kimi K3 support prompt caching?
Yes - both bill cached reads below their input rate, so warm-prefix workloads cost less than the list rates suggest. The exact cache-read rows are in the pricing table above.