Model market week of October 5, 2026

The market this week

New in the index (4):

• Ling 3.1 Flash (InclusionAI) — from Free/MTok input
• Pareto 26.10 Preview (Unbiased) — from $0.800/MTok input
• GPT-6.1 Sol Pro (OpenAI) — from $2.00/MTok input
• GPT-6.1 Sol (OpenAI) — from $2.00/MTok input

Price moves — largest 8 of 27 observed moves:

• DeepSeek V4 Pro 0813 input: $0.110 → $0.190/MTok (+72.8%)
• DeepSeek V4 Flash 0423 input: $0.030 → $0.0082/MTok (-72.7%)
• GLM 5.2 input: $0.180 → $0.289/MTok (+60.7%)
• GLM 5.3 input: $0.120 → $0.050/MTok (-58.3%)
• GLM 5.3 output: $1.14 → $0.488/MTok (-57.2%)
• DeepSeek V3.2 input: $0.134 → $0.209/MTok (+55.4%)
• DeepSeek V3.2 output: $0.202 → $0.310/MTok (+53.6%)
• DeepSeek V4 Flash 0423 output: $0.056 → $0.084/MTok (+50.0%)

Provider availability:

• DeepSeek V4.1 Flash: now on OpenInference, Decart, Reka, Io Net; left OpenInference, Io Net, Reka, krea
• GLM 5.3 Flash: now on PrimeIntellect, DekaLLM; left Io Net, PrimeIntellect
• Qwen3.8 27B: now on ModelRun, Cerebras; left CheaperInference
• DeepSeek V4 Pro 0813: now on Nebius, Sail Research, Reka; left Reka, Io Net
• DeepSeek V4 Flash 0731: now on Nebius; left Makora, Nebius
• Gemini 3.5 Flash Lite: left CheaperInference
• GLM 5.2: now on Nebius, Reka, InferenceNet, Morph, Reka; left Fireworks, Reka, Reka
• Kimi K2.7 Code: now on Tenstorrent; left Tenstorrent, DeepInfra

Most-used models (OpenRouter tokens, trailing 7 days):

• DeepSeek V4.1 Flash — 28.4T tokens
• GLM 5.3 Flash — 10.0T tokens
• MiMo-V2.6-Flash — 9.7T tokens
• Hy4 preview — 6.4T tokens
• GPT-6 Luna — 6.2T tokens

Measured note: x-ai/grok-4.6 on xai slowed week-over-week — median TTFT 1.25s → 2.01s (≥3 daily probes each week).

Around the industry

• OpenAI released GPT-6.1 Sol on September 29 at $2 per million input tokens and $10 per million output tokens, roughly one-fifth the token price of GPT-6 Astra. [3]
• OpenAI canceled the planned public release of GPT-6.1 Astra on September 28 after determining that the model frequently ignored instructions. [9]
• Google announced Gemini 4 Argon on September 30, initially rolling it out to trusted cyber defenders through its Fairwind Program. [4]
• Google launched Gemini 4 Argon at an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95%; regular pricing is $4 and $20, respectively. [4]

Assembled automatically from ModelIndex's observed data and cited external sources. Sourced claims link to their citations; our own figures link to the live pages they come from.