OCT 13, 2026 · MAPLEHILL LABS · WEEKLY · MARKET
Model market week of October 12, 2026
The market this week
New in the index (6):
- Muse Wonder (Meta) — from $4.00/MTok input
- Step 5 Preview (StepFun) — from $1.00/MTok input
- Ling 3.0 Flash Sante (InclusionAI) — from $0.042/MTok input
- Claude Haiku 5.5 (Anthropic) — from $0.100/MTok input
- Nano Banana 2.1 (Google) — from $1.50/MTok input
- Mistral Large 4 (Mistral) — from $0.680/MTok input
Price moves — largest 8 of 26 observed moves:
- DeepSeek V4 Flash 0423 input: $0.017 → $0.045/MTok (+160.1%)
- GLM 5.2 input: $0.030 → $0.070/MTok (+133.3%)
- DeepSeek V4.1 Flash input: $0.0067 → $0.0001/MTok (-98.5%)
- DeepSeek V4 Pro 0423 output: $1.90 → $0.341/MTok (-82.0%)
- DeepSeek V4 Pro 0813 output: $1.98 → $0.657/MTok (-66.8%)
- Qwen3.8 27B output: $1.35 → $0.500/MTok (-63.0%)
- Kimi K3 input: $0.450 → $0.730/MTok (+62.2%)
- Qwen3.8 27B input: $0.040 → $0.020/MTok (-50.0%)
Provider availability:
- DeepSeek V4.1 Flash: now on Cloudflare, Crusoe, OpenInference; left DekaLLM, NextBit, OpenInference
- Hy4 preview: now on Makora
- GLM 5.3: now on ModelRun, Tenstorrent, Io Net; left Nebius, Relace
- Qwen3.8 2.4T A95B: left DeepInfra
- DeepSeek V4 Pro 0813: now on Reka; left Nebius, Relace, NextBit
- GLM 5.3 Flash: left DekaLLM, Cloudflare, NextBit
- Qwen3.8 27B: now on Tenstorrent, Near AI
- DeepSeek V4 Pro 0423: left Relace, NextBit
Most-used models (OpenRouter tokens, trailing 7 days):
- DeepSeek V4.1 Flash — 40.5T tokens
- GLM 5.3 Flash — 12.7T tokens
- Step 5 Preview — 12.1T tokens
- MiMo-V2.6-Flash — 11.5T tokens
- Hy4 preview — 8.3T tokens
Measured note: z-ai/glm-5.3-flash on pareto-inference sped up week-over-week — median TTFT 3.02s → 1.27s (≥3 daily probes each week).
Around the industry
- Anthropic launched Claude Haiku 5.5 on October 7, describing it as its fastest small model and pricing it about 75% below Haiku 4.5 [1]
- Anthropic cut Claude Sonnet 5.5 cache-read pricing by 50% on October 7, making most agentic workloads about 20% cheaper [1]
- OpenAI began globally rolling out GPT-6 with Intelligent UI to ChatGPT Plus, Pro, Business, and Enterprise users on October 7 [2]
- OpenAI expanded the GPT-6 rollout to ChatGPT Free and Go users on October 8 [2]
- OpenAI made GPT-6.1 Sol ultrafast mode available to all API users on October 8, subject to rate limits [8]
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.