KaydedilenlerDerginizi ekleyin

Dergiler / International Journal of Multidisciplinary Studies and Innovative Technologies / 2026 / Cilt: 10 Sayı: 1

Büyük Dil Modeli Sağlayıcılarının Karşılaştırmalı Analizi: Performans, Ölçeklenebilirlik, Güvenilirlik ve Sürdürülebilirlik

A Comparative Analysis of Large Language Model Providers: Performance, Scalability, Reliability, and Sustainability

Khaled Hejja(ABDULLAH GUL UNIVERSITY, FACULTY OF ENGINEERING)0009-0003-9346-6618
Rifat Kurban(ABDULLAH GÜL ÜNİVERSİTESİ)0000-0002-0277-2210
Sayfa
46–56
Yayın Tarihi
2026-08-31
🏛️Bu makalenin bibliyografik verileri ve tam metin bağlantısı DergiPark üzerinden sağlanmıştır.
DergiPark Kaydı ↗
✨

Yapay Zeka Hızlı Bakış & Çıkarımlar

AI Insight
⚡ TL;DR(1 Cümlede Hızlı Çıkarım)

Ortaya çıkan çerçeve, sorumlu üretim dağıtımı için ön koşullar olarak göreve özgü değerlendirmeyi, kademeli model yönlendirmesini, açık yedekleme tasarımını ve sorgu başına kaynak muhasebesini vurgulamaktadır.

📌 3 Maddede Bu Makale

🎯
Amaç & Hipotez:Büyük dil modeli (LLM) sağlayıcı seçimi, basit bir kıyaslama sıralaması egzersizinden ziyade çok amaçlı bir altyapı kararı haline gelmiştir.
📊
Temel Bulgular & Veri:Kıyaslama kanıtlarını, sistem literatürünü, sağlayıcı açıklamalarını, kesinti verilerini ve çevresel çalışmaları sentezleyerek üç ortak örüntüyü ortaya koymaktadır: kirlenmeden arındırılmış ve gerçek dünya değerlendirmesi altında kıyaslama liderliği giderek daha kırılgan hale gelmektedir; seyrek ve optimize edilmiş sunucu mimarileri, öncü ve açık kaynaklı sistemler arasındaki maliyet farkını azaltmaktadır; ve güvenilirlik ve sürdürülebilirlik açıklamaları kurumsal benimsemenin gerisinde kalmaktadır.
💡
Sonuç & Katkı:Ortaya çıkan çerçeve, sorumlu üretim dağıtımı için ön koşullar olarak göreve özgü değerlendirmeyi, kademeli model yönlendirmesini, açık yedekleme tasarımını ve sorgu başına kaynak muhasebesini vurgulamaktadır.
🔬Tür:Uygulamalı Anket & Alan Araştırması
📍Kapsam:Genişletilmiş Akademik Örneklem
⚙️Değişkenler:Büyük dil modelleri · LLM sağlayıcıları · karşılaştırmalı analiz

Özet

Büyük dil modeli (LLM) sağlayıcı seçimi, basit bir kıyaslama sıralaması egzersizinden ziyade çok amaçlı bir altyapı kararı haline gelmiştir. Bu makale, başlıca kapalı, açık kaynaklı ve hiper ölçekli LLM seçeneklerini dört boyutta karşılaştırmaktadır: performans, ölçeklenebilirlik, güvenilirlik ve sürdürülebilirlik. Kıyaslama kanıtlarını, sistem literatürünü, sağlayıcı açıklamalarını, kesinti verilerini ve çevresel çalışmaları sentezleyerek üç ortak örüntüyü ortaya koymaktadır: kirlenmeden arındırılmış ve gerçek dünya değerlendirmesi altında kıyaslama liderliği giderek daha kırılgan hale gelmektedir; seyrek ve optimize edilmiş sunucu mimarileri, öncü ve açık kaynaklı sistemler arasındaki maliyet farkını azaltmaktadır; ve güvenilirlik ve sürdürülebilirlik açıklamaları kurumsal benimsemenin gerisinde kalmaktadır. Ortaya çıkan çerçeve, sorumlu üretim dağıtımı için ön koşullar olarak göreve özgü değerlendirmeyi, kademeli model yönlendirmesini, açık yedekleme tasarımını ve sorgu başına kaynak muhasebesini vurgulamaktadır.

Abstract

Large language model (LLM) provider selection has become a multi-objective infrastructure decision rather than a simple benchmark ranking exercise. This paper compares major proprietary, open-weight, and hyperscaler-mediated LLM options across four dimensions: performance, scalability, reliability, and sustainability. It synthesizes benchmark evidence, systems literature, provider disclosures, outage data, and environmental studies to show three cross-cutting patterns: benchmark leadership is increasingly fragile under contamination-free and real-world evaluation; sparse and optimized serving architectures are compressing the cost gap between frontier and open-weight systems; and reliability and sustainability disclosures lag behind enterprise adoption. The resulting framework emphasizes task-specific evaluation, tiered model routing, explicit fallback design, and per-query resource accounting as prerequisites for responsible production deployment.

🔖

Bu Makaleye Atıf Yap

Hejja, K., Kurban, R. (2026). Büyük Dil Modeli Sağlayıcılarının Karşılaştırmalı Analizi: Performans, Ölçeklenebilirlik, Güvenilirlik ve Sürdürülebilirlik. International Journal of Multidisciplinary Studies and Innovative Technologies, Cilt: 10 Sayı: 1, 46–56. https://doi.org/10.36287/ijmsit.10.1.5

📄 Tam Metin PDF Okuyucu

📄

PDF görüntüleyici hazırlanıyor...

Kaynakça (60)

  1. OpenAI, "The next phase of the Microsoft OpenAI partnership," Apr. 27, 2026. https://openai.com/index/next-phase-of-microsoft-partnership/ (accessed May 25, 2026).
  2. DeepSeek-AI, "DeepSeek-V3 technical report," arXiv preprint arXiv:2412.19437, 2024. https://doi.org/10.48550/arXiv.2412.19437
  3. W. Kwon et al., "Efficient memory management for large language model serving with PagedAttention," in Proc. SOSP, 2023, pp. 611-626. https://doi.org/10.1145/3600006.3613165
  4. L. Zheng et al., "SGLang: Efficient execution of structured language model programs," in Proc. NeurIPS, 2024. https://doi.org/10.48550/arXiv.2312.07104
  5. N. Jegham, M. Abdelatti, C. Y. Koh, L. Elmoubarki, and A. Hendawi, "How hungry is AI? Benchmarking energy, water, and carbon footprint of LLM inference," arXiv preprint arXiv:2505.09598, 2025. https://doi.org/10.48550/arXiv.2505.09598
  6. OpenAI, "Why SWE-bench Verified no longer measures frontier coding capabilities," OpenAI Blog, Feb. 23, 2026. https://openai.com/index/why-we-no-longer-evaluate-swe-bench-verified/ (accessed May 25, 2026).
  7. Q. Zhao, Y. Huang, T. Lv, L. Cui, et al., "MMLU-CF: A contamination-free multi-task language understanding benchmark," in Proc. ACL, 2025, pp. 13371-13391. https://doi.org/10.18653/v1/2025.acl-long.656
  8. F. Yao, Y. Zhuang, Z. Sun, S. Xu, A. Kumar, and J. Shang, "Data contamination can cross language barriers," in Proc. EMNLP, 2024, pp. 17864-17875. https://doi.org/10.18653/v1/2024.emnlp-main.990
  9. N. Jain, K. Han, A. Gu, W.-D. Li, F. Yan, et al., "LiveCodeBench: Holistic and contamination free evaluation of large language models for code," in Proc. ICLR, 2025. https://doi.org/10.48550/arXiv.2403.07974
  10. J. Kaplan et al., "Scaling laws for neural language models," arXiv preprint arXiv:2001.08361, 2020. https://doi.org/10.48550/arXiv.2001.08361
  11. J. Hoffmann et al., "Training compute-optimal large language models," in Proc. NeurIPS, 2022. https://doi.org/10.48550/arXiv.2203.15556
  12. A. Chowdhery et al., "PaLM: Scaling language modeling with pathways," Journal of Machine Learning Research, vol. 24, no. 240, pp. 1-113, 2023. https://doi.org/10.48550/arXiv.2204.02311
  13. W. Fedus, B. Zoph, and N. Shazeer, "Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity," Journal of Machine Learning Research, vol. 23, no. 120, pp. 1-39, 2022. https://doi.org/10.48550/arXiv.2101.03961
  14. T. B. Brown et al., "Language models are few-shot learners," in Proc. NeurIPS, 2020, pp. 1877-1901. https://doi.org/10.48550/arXiv.2005.14165
  15. A. Grattafiori et al., "The Llama 3 herd of models," arXiv preprint arXiv:2407.21783, 2024. https://doi.org/10.48550/arXiv.2407.21783
  16. A. Yang, A. Li, B. Yang, et al., "Qwen3 technical report," arXiv preprint arXiv:2505.09388, 2025. https://doi.org/10.48550/arXiv.2505.09388
  17. D. Hendrycks et al., "Measuring massive multitask language understanding," in Proc. ICLR, 2021. https://doi.org/10.48550/arXiv.2009.03300
  18. Y. Wang et al., "MMLU-Pro: A more robust and challenging multi-task language understanding benchmark," in Proc. NeurIPS, 2024. https://doi.org/10.48550/arXiv.2406.01574
  19. M. Chen et al., "Evaluating large language models trained on code," arXiv preprint arXiv:2107.03374, 2021. https://doi.org/10.48550/arXiv.2107.03374
  20. C. E. Jimenez et al., "SWE-bench: Can language models resolve real-world GitHub issues?" in Proc. ICLR, 2024. https://doi.org/10.48550/arXiv.2310.06770
  21. M. Rahman, S. Khatoonabadi, and E. Shihab, "Beyond synthetic benchmarks: Evaluating LLM performance on real-world class-level code generation," arXiv preprint arXiv:2510.26130, 2025. https://doi.org/10.48550/arXiv.2510.26130
  22. S. Liang, Y. Hu, N. Jiang, and L. Tan, "Can language models replace programmers for coding? REPOCOD says 'not yet'," in Proc. ACL, 2025, pp. 24698-24717. https://doi.org/10.18653/v1/2025.acl-long.1204
  23. E. J. Gong, C. S. Bang, J. J. Lee, et al., "Knowledge-practice performance gap in clinical large language models: Systematic review of 39 benchmarks," Journal of Medical Internet Research, vol. 27, e84120, 2025. https://doi.org/10.2196/84120
  24. T. Xie et al., "OSWorld: Benchmarking multimodal agents for open-ended tasks in real computer environments," in Proc. NeurIPS, 2024. https://doi.org/10.48550/arXiv.2404.07972
  25. T. Dao et al., "FlashAttention: Fast and memory-efficient exact attention with IO-awareness," in Proc. NeurIPS, 2022, pp. 16344-16359. https://doi.org/10.48550/arXiv.2205.14135
  26. J. Shah, G. Bikshandi, Y. Zhang, V. Thakkar, P. Ramani, and T. Dao, "FlashAttention-3: Fast and accurate attention with asynchrony and low-precision," in Proc. NeurIPS, 2024. https://doi.org/10.48550/arXiv.2407.08608
  27. E. Frantar et al., "GPTQ: Accurate post-training quantization for generative pre-trained transformers," in arXiv preprint, 2023. https://doi.org/10.48550/arXiv.2210.17323
  28. J. Lin et al., "AWQ: Activation-aware weight quantization for LLM compression and acceleration," in Proc. MLSys, 2024. https://doi.org/10.48550/arXiv.2306.00978
  29. G. Xiao, J. Lin, M. Seznec, H. Wu, J. Demouth, and S. Han, "SmoothQuant: Accurate and efficient post-training quantization for large language models," in Proc. PMLR, pp. 38087-38099, 2023. https://doi.org/10.48550/arXiv.2211.10438
  30. Y. Lin et al., "QServe: W4A8KV4 quantization and system co-design for efficient LLM serving," in Proc. MLSys, 2025. https://doi.org/10.48550/arXiv.2405.04532
  31. E. Strubell, A. Ganesh, and A. McCallum, "Energy and policy considerations for deep learning in NLP," in Proc. ACL, 2019, pp. 3645-3650. https://doi.org/10.18653/v1/P19-1355
  32. D. Patterson et al., "Carbon emissions and large neural network training," arXiv preprint arXiv:2104.10350, 2021. https://doi.org/10.48550/arXiv.2104.10350
  33. M. Dahl, V. Magesh, M. Suzgun, and D. E. Ho, "Large legal fictions: Profiling legal hallucinations in large language models," Journal of Legal Analysis, vol. 16, no. 1, pp. 64-93, 2024. https://doi.org/10.1093/jla/laae003
  34. X. Chu, S. Talluri, Q. Lu, and A. Iosup, "An empirical characterization of outages and incidents in public services for large language models," in Proc. ICPE, 2025, pp. 69-80. https://doi.org/10.1145/3676151.3719372
  35. B. Atil, S. Aykent, A. Chittams, et al., "Non-determinism of 'deterministic' LLM system settings in hosted environments," in Proc. Eval4NLP, 2025, pp. 135-148. https://doi.org/10.18653/v1/2025.eval4nlp-1.12
  36. OpenAI, "Scale Tier for API customers," 2026. https://openai.com/api-scale-tier/ (accessed May 25, 2026).
  37. Microsoft, "Service Level Agreements (SLA) for Online Services," 2026, and "Provisioned throughput unit (PTU) costs and billing," Microsoft Learn, 2026. https://www.microsoft.com/licensing/docs/view/Service-Level-Agreements-SLA-for-Online-Services; https://learn.microsoft.com/en-us/azure/foundry/openai/how-to/provisioned-throughput-onboarding (accessed May 25, 2026).
  38. Anthropic, "Service tiers," Claude API Docs, 2026. https://platform.claude.com/docs/en/api/service-tiers (accessed May 25, 2026).
  39. Amazon Web Services, "Amazon Bedrock Service Level Agreement," 2023. https://aws.amazon.com/bedrock/sla/ (accessed May 25, 2026).
  40. OpenAI, "GPT-5 is here," 2026. https://openai.com/gpt-5/ (accessed May 25, 2026).
  41. Anthropic, "Introducing Claude Opus 4.7," Apr. 16, 2026. https://www.anthropic.com/news/claude-opus-4-7 (accessed May 25, 2026).
  42. Google, "Gemini 3.1 Pro: A smarter model for your most complex tasks," Feb. 19, 2026, and "Building the agentic future: Developer highlights from I/O 2026," May 19, 2026. https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/; https://blog.google/innovation-and-ai/technology/developers-tools/google-io-2026-developer-highlights/ (accessed May 25, 2026).
  43. xAI, "Grok Model Retirement on May 15, 2026," xAI Docs, 2026. https://docs.x.ai/developers/migration/may-15-retirement (accessed May 25, 2026).
  44. Alibaba Group, "Alibaba open-sources Qwen3.5, a natively multimodal model built for high-efficiency inference," Feb. 16, 2026. https://www.alibabagroup.com/en-US/document-1960233590314762240 (accessed May 25, 2026).
  45. Qwen Team, "Qwen3.6," GitHub repository, 2026. https://github.com/QwenLM/Qwen3.6 (accessed May 25, 2026).
  46. Kimi Team, "Kimi K2.5: Visual agentic intelligence," arXiv preprint arXiv:2602.02276, 2026. https://doi.org/10.48550/arXiv.2602.02276
  47. Moonshot AI, "Kimi-K2.6 model card," Hugging Face, 2026. https://huggingface.co/moonshotai/Kimi-K2.6 (accessed May 25, 2026).
  48. DeepSeek-AI, A. Liu, A. Mei, et al., "DeepSeek-V3.2: Pushing the frontier of open large language models," arXiv preprint arXiv:2512.02556, 2025. https://doi.org/10.48550/arXiv.2512.02556
  49. DeepSeek-AI, "DeepSeek V4 model card," Model Transparency Documentation, Apr. 24, 2026. https://fe-static.deepseek.com/chat/transparency/deepseek-V4-model-card-EN.pdf (accessed May 26, 2026).
  50. Groq, "Groq LPU inference engine leads in first independent LLM benchmark," Feb. 13, 2024, and "LPU architecture," Groq, 2025. https://groq.com/newsroom/groq-lpu-inference-engine-leads-in-first-independent-llm-benchmark; https://groq.com/lpu-architecture (accessed May 25, 2026).
  51. N. Shazeer et al., "Outrageously large neural networks: The sparsely-gated mixture-of-experts layer," in arXiv preprint, 2017. https://doi.org/10.48550/arXiv.1701.06538
  52. A. Q. Jiang et al., "Mixtral of experts," arXiv preprint arXiv:2401.04088, 2024. https://doi.org/10.48550/arXiv.2401.04088
  53. T. Dao, "FlashAttention-2: Faster attention with better parallelism and work partitioning," in Proc. ICLR, 2024. https://doi.org/10.48550/arXiv.2307.08691
  54. Cloud provider GPU pricing pages: AWS EC2 On-Demand and P5 instance pricing, Microsoft Azure NCads H100 v5 pricing, and Google Cloud GPU pricing, 2025-2026. https://aws.amazon.com/ec2/pricing/on-demand/; https://aws.amazon.com/ec2/instance-types/p5/; https://learn.microsoft.com/en-us/azure/virtual-machines/sizes/gpu-accelerated/ncadsh100v5-series; https://cloud.google.com/compute/gpus-pricing (accessed May 25, 2026).
  55. P. Melendez and M. Dalida, "Don't let resource exhaustion leave your users hanging: A guide to handling 429 errors," Google Cloud Blog, Nov. 22, 2024. https://cloud.google.com/blog/products/ai-machine-learning/learn-how-to-handle-429-resource-exhaustion-errors-in-your-llms (accessed May 25, 2026).
  56. A. S. Luccioni, S. Viguier, and A.-L. Ligozat, "Estimating the carbon footprint of BLOOM, a 176B parameter language model," Journal of Machine Learning Research, vol. 24, no. 253, pp. 1-15, 2023. https://doi.org/10.48550/arXiv.2211.02001
  57. J. Dodge et al., "Measuring the carbon intensity of AI in cloud instances," in Proc. FAccT, 2022, pp. 1877-1894. https://doi.org/10.1145/3531146.3533234
  58. K. Crawford, "Generative AI's environmental costs are soaring and mostly secret," Nature, vol. 626, p. 693, 2024. https://doi.org/10.1038/d41586-024-00478-x
  59. European Parliament and Council of the European Union, "Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act)," Official Journal of the European Union, L 2024/1689, 2024. https://eur-lex.europa.eu/eli/reg/2024/1689
  60. ISO/IEC, "Information technology - Software Carbon Intensity (SCI) specification (ISO/IEC 21031:2024)," International Organization for Standardization, 2024. https://www.iso.org/standard/86612.html