DEV Community: Jemmmm

GLM-5.2 vs Claude Fable 5: ngân sách đầu ra, reasoning_tokens và mức discount 0.8

Jemmmm — Mon, 06 Jul 2026 16:11:30 +0000

GLM-5.2 vs Claude Fable 5: khác biệt lớn nằm ở ngân sách đầu ra

Đây không phải là một bảng xếp hạng mô hình chung chung. Trong lần gọi API thực tế này, GLM-5.2 giải đúng các bài suy luận sau khi tăng ngân sách đầu ra, nhưng ở ngân sách thấp phần nội dung nhìn thấy có thể rỗng. Claude Fable 5 ổn định hơn với ngân sách thấp và tốt hơn trong bài tạo HTML animation dài.

Vì sao bài test này đáng chú ý

The test used the Crazyrouter OpenAI-compatible API rather than a chat UI. That matters because the result was not judged only by prose quality. Each response was checked with operational metadata:

Base URL: https://cn.crazyrouter.com/v1
Endpoint: POST /v1/chat/completions
Models: glm-5.2, claude-fable-5
temperature: 0.2
Test date: 2026-07-06

The important fields were max_tokens, completion_tokens, reasoning_tokens, finish_reason, visible content length, whether the generated HTML was closed, and whether the animation actually moved in a browser.

Thiết kế bài test

The benchmark deliberately mixed three task types:

Task	Purpose	Reference result
`MATH-003`	State-based expectation reasoning	Expected flips until HH = `6`
`PHYS-003`	Momentum plus energy accounting	`V = 3.0 m/s`, `x ≈ 0.148 m`
`CODE-003-ANIM`	Long runnable artifact generation	Complete 800x500 Canvas animation HTML

The first two tasks measured reasoning. The third task measured whether a model can produce a complete artifact, not merely a convincing partial code block.

Kết quả quan sát

Task	`glm-5.2`	`claude-fable-5`
Math, original budget	`finish_reason=length`, `completion_tokens=1601`, `reasoning_tokens=1600`, visible body empty	`finish_reason=stop`, complete and correct
Math, retest	Correct after `max_tokens=3200`	Retest not needed
Physics, original budget	`finish_reason=length`, visible body empty	Complete and correct
Physics, retest	Correct after `max_tokens=8000`	Retest not needed
Animation, original budget	Empty visible HTML at `max_tokens=3200`	Partial HTML, truncated
Animation, retest	Still truncated at `max_tokens=8000`	Complete HTML; browser validation passed

The most important observation is that GLM-5.2 was not failing the reasoning itself. In the math and physics tasks, it produced correct answers after a larger output budget. The problem was visibility and completion: a request could return HTTP 200 while the user-facing content was empty or incomplete.

For the long Canvas animation, the difference was sharper. GLM-5.2 produced a visible HTML fragment at max_tokens=8000, but it stopped inside JavaScript and did not close the file. Claude Fable 5 completed the HTML at max_tokens=8000; browser validation showed no console errors, an 800x500 canvas, controls, a speed slider, and changedPixels=55090 after 700 ms.

Góc nhìn chi phí

Tại thời điểm viết bài, pricing API của Crazyrouter trả về discount: 0.8 cho glm-5.2. Vì vậy GLM-5.2 có lợi thế chi phí rõ ràng nếu bạn theo dõi kỹ reasoning_tokens, finish_reason và đặt max_tokens phù hợp.

This is the practical tradeoff:

Workload	Better fit from this test
Short reasoning with enough output budget	GLM-5.2 can be a cost-effective option
Low-budget reasoning responses	Claude Fable 5 was steadier
Long single-file code generation	Claude Fable 5 was stronger in this run
Batch evaluations where metadata is logged	GLM-5.2 becomes easier to operate safely

Do not treat the 0.8 multiplier as a permanent universal price. It is a pricing-data snapshot from Crazyrouter at publication time and should be checked again before a large deployment.

Ghi chú tích hợp

Minimal request:

curl https://cn.crazyrouter.com/v1/chat/completions \
  -H "Authorization: Bearer $CRAZYROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.2",
    "messages": [
      {
        "role": "user",
        "content": "Solve the HH expected-flips problem with state equations."
      }
    ],
    "temperature": 0.2,
    "max_tokens": 3200
  }'

To compare Claude Fable 5, keep the same payload and change only the model:

{
  "model": "claude-fable-5"
}

For production-style evaluations, log this shape for every request:

{
  "model": "glm-5.2",
  "max_tokens": 3200,
  "finish_reason": "length",
  "completion_tokens": 3200,
  "reasoning_tokens": 3178,
  "visible_content_chars": 0,
  "html_closed": false,
  "browser_validation": "not_run_incomplete_html"
}

API endpoints should stay clean. Do not add UTM parameters to https://cn.crazyrouter.com/v1. Use tracking only on human-facing article or registration links.

Bạn có thể chạy cùng kiểu request OpenAI-compatible trên Crazyrouter để so sánh hai mô hình bằng prompt thật của mình.

https://crazyrouter.com/register?utm_source=devto&utm_medium=article&utm_campaign=glm52_fable5_budget_cost_20260706&utm_content=devto_glm-52-vs-claude-fable-5-output-budget-cost-vi_20260706__bottom&utm_term=glm-5.2+claude+fable+5+benchmark

FAQ

Did GLM-5.2 fail the reasoning tasks?

No. In this run, GLM-5.2 solved the math task after max_tokens=3200 and the physics task after max_tokens=8000. The issue was that lower budgets were consumed mostly by reasoning tokens before visible content appeared.

Why not score HTTP 200 as success?

Because HTTP 200 only means the API call returned. A benchmark answer can still be unusable if finish_reason=length, visible content is empty, or generated code is incomplete.

Why was the animation task included?

Long code generation exposes a different failure mode. A model can write a convincing first half of a file and still fail if the HTML or JavaScript is cut off.

Is GLM-5.2 still worth testing?

Yes. The current 0.8 discount multiplier makes it attractive for workloads where you can allocate enough output budget and monitor response metadata.

What should be recorded in future comparisons?

At minimum: max_tokens, completion_tokens, reasoning_tokens, finish_reason, visible output length, artifact completeness, and runtime validation.

Final verdict

Kết luận thực tế: GLM-5.2 hấp dẫn về chi phí và có thể suy luận tốt, nhưng cần kiểm soát ngân sách đầu ra. Claude Fable 5 ổn định hơn cho câu trả lời ngắn và file HTML hoàn chỉnh trong lần thử này.

GLM-5.2 と Claude Fable 5 を実測比較：出力予算、reasoning_tokens、そして 0.8 価格係数

Jemmmm — Mon, 06 Jul 2026 16:10:52 +0000

GLM-5.2 vs Claude Fable 5：差が出たのは推論力だけではなく出力予算だった

この比較は「どちらが絶対に強いか」を決める記事ではありません。実際の API 呼び出しで、GLM-5.2 は出力予算を増やすと数学・物理の推論を正しく返しました。一方で Claude Fable 5 は低い予算でも短く安定し、長い HTML アニメーションではより確実に完走しました。

このテストを見る理由

The test used the Crazyrouter OpenAI-compatible API rather than a chat UI. That matters because the result was not judged only by prose quality. Each response was checked with operational metadata:

Base URL: https://cn.crazyrouter.com/v1
Endpoint: POST /v1/chat/completions
Models: glm-5.2, claude-fable-5
temperature: 0.2
Test date: 2026-07-06

テストした課題

The benchmark deliberately mixed three task types:

Task	Purpose	Reference result
`MATH-003`	State-based expectation reasoning	Expected flips until HH = `6`
`PHYS-003`	Momentum plus energy accounting	`V = 3.0 m/s`, `x ≈ 0.148 m`
`CODE-003-ANIM`	Long runnable artifact generation	Complete 800x500 Canvas animation HTML

The first two tasks measured reasoning. The third task measured whether a model can produce a complete artifact, not merely a convincing partial code block.

観測結果

Task	`glm-5.2`	`claude-fable-5`
Math, original budget	`finish_reason=length`, `completion_tokens=1601`, `reasoning_tokens=1600`, visible body empty	`finish_reason=stop`, complete and correct
Math, retest	Correct after `max_tokens=3200`	Retest not needed
Physics, original budget	`finish_reason=length`, visible body empty	Complete and correct
Physics, retest	Correct after `max_tokens=8000`	Retest not needed
Animation, original budget	Empty visible HTML at `max_tokens=3200`	Partial HTML, truncated
Animation, retest	Still truncated at `max_tokens=8000`	Complete HTML; browser validation passed

費用対効果の見方

執筆時点で Crazyrouter の pricing API は glm-5.2 に discount: 0.8 を返しています。つまり、reasoning_tokens と max_tokens をきちんと監視できる用途では、GLM-5.2 はかなり費用対効果の高い選択肢になります。

This is the practical tradeoff:

Workload	Better fit from this test
Short reasoning with enough output budget	GLM-5.2 can be a cost-effective option
Low-budget reasoning responses	Claude Fable 5 was steadier
Long single-file code generation	Claude Fable 5 was stronger in this run
Batch evaluations where metadata is logged	GLM-5.2 becomes easier to operate safely

Do not treat the 0.8 multiplier as a permanent universal price. It is a pricing-data snapshot from Crazyrouter at publication time and should be checked again before a large deployment.

実装時の注意

Minimal request:

curl https://cn.crazyrouter.com/v1/chat/completions \
  -H "Authorization: Bearer $CRAZYROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.2",
    "messages": [
      {
        "role": "user",
        "content": "Solve the HH expected-flips problem with state equations."
      }
    ],
    "temperature": 0.2,
    "max_tokens": 3200
  }'

To compare Claude Fable 5, keep the same payload and change only the model:

{
  "model": "claude-fable-5"
}

For production-style evaluations, log this shape for every request:

{
  "model": "glm-5.2",
  "max_tokens": 3200,
  "finish_reason": "length",
  "completion_tokens": 3200,
  "reasoning_tokens": 3178,
  "visible_content_chars": 0,
  "html_closed": false,
  "browser_validation": "not_run_incomplete_html"
}

API endpoints should stay clean. Do not add UTM parameters to https://cn.crazyrouter.com/v1. Use tracking only on human-facing article or registration links.

同じ OpenAI 互換リクエストを Crazyrouter で流し、自分のプロンプトで両モデルを比較できます。

https://crazyrouter.com/register?utm_source=devto&utm_medium=article&utm_campaign=glm52_fable5_budget_cost_20260706&utm_content=devto_glm-52-vs-claude-fable-5-output-budget-cost-ja_20260706__bottom&utm_term=glm-5.2+claude+fable+5+benchmark

FAQ

Did GLM-5.2 fail the reasoning tasks?

Why not score HTTP 200 as success?

Because HTTP 200 only means the API call returned. A benchmark answer can still be unusable if finish_reason=length, visible content is empty, or generated code is incomplete.

Why was the animation task included?

Long code generation exposes a different failure mode. A model can write a convincing first half of a file and still fail if the HTML or JavaScript is cut off.

Is GLM-5.2 still worth testing?

Yes. The current 0.8 discount multiplier makes it attractive for workloads where you can allocate enough output budget and monitor response metadata.

What should be recorded in future comparisons?

At minimum: max_tokens, completion_tokens, reasoning_tokens, finish_reason, visible output length, artifact completeness, and runtime validation.

Final verdict

結論は単純ではありません。GLM-5.2 はコスト面で魅力があり推論も可能ですが、出力予算の管理が必要です。Claude Fable 5 は短い回答と完成した単一 HTML 生成で安定していました。

Suno SDK 시리즈 03: generate, extend, cover, 가사, 보컬 분리를 운영 환경에 연결하는 순서

Jemmmm — Mon, 06 Jul 2026 07:59:52 +0000

Suno SDK 시리즈 03: generate, extend, cover, 가사, 보컬 분리를 운영 환경에 연결하는 순서

Crazyrouter의 Suno SDK는 하나의 모델이 아니라 여러 음악 생성 및 편집 작업을 묶은 모델 패밀리로 보는 것이 자연스럽습니다.

주요 모델은 다음과 같습니다.

suno/generate
suno/extend
suno/upload-extend
suno/cover
suno/timestamped-lyrics
suno/vocal-separation

운영 환경에서는 처음부터 모든 기능을 UI에 노출하지 않는 편이 안전합니다. 먼저 suno/generate를 안정화하고, 그 결과에서 audio_url과 작업 ID를 안정적으로 얻을 수 있는지 확인한 뒤 파생 기능을 열어야 합니다.

권장 구현 순서

1. suno/generate
2. 생성 결과 audio_url 다운로드 검증
3. audioId / audio_url / task_id 저장
4. suno/extend
5. suno/upload-extend
6. suno/cover
7. timestamped-lyrics / vocal-separation

후속 기능은 이전 생성 결과에 의존하는 경우가 많습니다. generate 결과에 안정적인 음원 URL이나 식별자가 없다면 extend나 cover 버튼을 보여주는 것은 실패율만 높입니다.

버전 지정

Suno 통합 작업에서는 버전을 model_version으로 지정합니다.

V5_5
V5
V4_5PLUS
V4_5ALL
V4_5
V4

예전 라우트에서 사용하던 mv와 혼동하지 않아야 합니다.

generate의 역할

suno/generate는 가장 먼저 연결해야 할 기본 기능입니다.

사용 예:

가사로 노래 생성
음악 스타일 설명으로 데모곡 생성
광고나 영상용 짧은 음악 생성
제품 내부 샘플 음원 생성

최소 테스트에서는 작업 제출과 폴링뿐 아니라 resultJson.data[0].audio_url 다운로드까지 확인해야 합니다.

extend와 cover의 주의점

extend는 기존 음원을 이어 붙이는 기능, cover는 기존 음원을 다른 스타일로 변환하는 기능으로 이해하면 쉽습니다.

두 기능은 보통 다음 정보에 의존합니다.

원본 audio URL
원본 task_id
상위 서비스가 반환하는 audioId 계열 식별자

이 정보가 없다면 프론트엔드에서 버튼을 보여주지 않는 편이 낫습니다.

timestamped-lyrics와 vocal-separation

시간 정보가 있는 가사와 보컬 분리는 생성 본체보다 결과 필드 의존도가 더 높습니다. 첫 릴리스에서는 선택 기능으로 두고, 실제 데이터로 충분히 검증한 뒤 공개하는 것을 권장합니다.

운영 체크리스트

체크	통과 조건
제출	`task_id` 반환
폴링	종료 상태를 정확히 처리
다운로드	`audio_url`이 HTTP 200 반환
파일 검증	`audio/*`이고 충분한 크기
파생 기능	필요한 ID와 URL 저장
실패 분류	`task_failed`와 `content_unavailable` 분리

정리

Suno SDK는 먼저 generate를 안정화하는 것부터 시작해야 합니다. 작업 성공만으로는 부족하며, audio_url 획득, 다운로드, 저장까지가 최소 운영 단위입니다.

그 다음 extend, cover, 시간 정보 가사, 보컬 분리를 단계적으로 열어야 사고가 적습니다.